ATP-Prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ATP Tennis Match Prediction  \n",
    "\n",
    "<img src=\"logoATP.png\"  width=\"300\">\n",
    "\n",
    "## Overview  \n",
    "In this notebook, we will try to conduct a full Data Science & ML Study of the ATP dataset to predict the outcome of the matches in the year 2017. The study will include : \n",
    "- Some exploratory data analysis through questions/answers\n",
    "- The data preparation process (Missing values imputation..)\n",
    "- The modeling phase (Feature Engineering, Model Tuning..)\n",
    "\n",
    "## DataSet \n",
    "The website http://tennis-data.co.uk/alldata.php gathers outcomes of ATP (Association of Tennis Professionals - men only) tennis games over several years (From year 2000). \n",
    "A short description of each variable can be found here : http://www.tennis-data.co.uk/notes.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from urllib.request import urlopen  \n",
    "import os.path as osp\n",
    "import os\n",
    "import logging\n",
    "import zipfile\n",
    "from glob import glob\n",
    "logging.getLogger().setLevel('INFO')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Helpers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def download_file(url_str, path):\n",
    "    url = urlopen(url_str)\n",
    "    output = open(path, 'wb')       \n",
    "    output.write(url.read())\n",
    "    output.close()  \n",
    "    \n",
    "def extract_file(archive_path, target_dir):\n",
    "    zip_file = zipfile.ZipFile(archive_path, 'r')\n",
    "    zip_file.extractall(target_dir)\n",
    "    zip_file.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Download the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2000/2000.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2001/2001.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2002/2002.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2003/2003.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2004/2004.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2005/2005.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2006/2006.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2007/2007.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2008/2008.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2009/2009.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2010/2010.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2011/2011.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2012/2012.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2013/2013.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2014/2014.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2015/2015.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2016/2016.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2017/2017.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2018/2018.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2000/2000.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2001/2001.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2002/2002.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2003/2003.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2004/2004.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2005/2005.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2006/2006.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2007/2007.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2008/2008.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2009/2009.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2010/2010.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2011/2011.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2012/2012.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2013/2013.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2014/2014.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2015/2015.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2016/2016.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2017/2017.zip\n",
      "INFO:root:downloading & extracting file http://tennis-data.co.uk/2018/2018.zip\n",
      "/Users/Ouadie/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:18: FutureWarning: Sorting because non-concatenation axis is not aligned. A future version\n",
      "of pandas will change to not sort by default.\n",
      "\n",
      "To accept the future behavior, pass 'sort=True'.\n",
      "\n",
      "To retain the current behavior and silence the warning, pass sort=False\n",
      "\n",
      "INFO:root:52236 matches ATP in df_atp\n"
     ]
    }
   ],
   "source": [
    "BASE_URL = 'http://tennis-data.co.uk'\n",
    "DATA_DIR = \"tennis_data\"\n",
    "ATP_DIR = './{}/ATP'.format(DATA_DIR)\n",
    "\n",
    "ATP_URLS = [BASE_URL + \"/%i/%i.zip\" % (i,i) for i in range(2000,2019)]\n",
    "\n",
    "os.makedirs(osp.join(ATP_DIR, 'archives'), exist_ok=True)\n",
    "\n",
    "for files, directory in ((ATP_URLS,ATP_DIR),(ATP_URLS,ATP_DIR)):\n",
    "    for dl_path in files:\n",
    "        logging.info(\"downloading & extracting file %s\", dl_path)\n",
    "        archive_path = osp.join(directory, 'archives', osp.basename(dl_path))\n",
    "        download_file(dl_path, archive_path)\n",
    "        extract_file(archive_path, directory)\n",
    "    \n",
    "ATP_FILES = sorted(glob(\"%s/*.xls*\" % ATP_DIR))\n",
    "\n",
    "df_atp = pd.concat([pd.read_excel(f) for f in ATP_FILES], ignore_index=True)\n",
    "\n",
    "logging.info(\"%i matches ATP in df_atp\", df_atp.shape[0])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploratory Data Analysis through questions  \n",
    "In this section, we will try to ask and respond to some questions to get used to the problem and gain some insights."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import re    #Regular Expression module for string manipulation\n",
    "import plotnine as pln #Plotting library (implementation of R's ggplot2)\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "sns.set(style=\"darkgrid\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Q1) **Who are the three ATP players with the most wins ?**  \n",
    "A quick exploration of the \"Winner\" feature in our data shows that names take the form **'XXX Y.'**. Still, some observations have an aditional space **'XXX Y. '**. So the first thing to do is to correct these exceptions.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "reCorr1 = lambda s: re.sub(string = s, pattern=\" $\", repl=\"\") "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp.loc[:,'Winner'] = df_atp.loc[:,'Winner'].apply(reCorr1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "winnersCount = df_atp['Winner'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Federer R.     1115\n",
       "Nadal R.        891\n",
       "Djokovic N.     799\n",
       "Name: Winner, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "winnersCount.sort_values(ascending=False).head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So the players with the most wins are :  \n",
    "1 - Federer R.  \n",
    "2 - Nadal R.      \n",
    "2 - Djokovic N.  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---------\n",
    "### Q2) **How many sets did the player “​Federer R.” win in total ?**  \n",
    "\n",
    "1- First, we filter the observations that include our player 'Federer R.', whether he's the winner or the loser.   \n",
    "2- We select the features we're going to use for this question : 'Wsets','Lsets','Winner'.  \n",
    "3- We check/clean the 'Wsets' & 'Lsets' features and turn their values into numerical values.  \n",
    "4- Finally, we can do the summation of the number of won sets.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Subset of Matches having 'Federer R.' as a player\n",
    "FedIndexes = (df_atp.Winner=='Federer R.')| (df_atp.Loser=='Federer R.')\n",
    "FedererSub = df_atp.loc[FedIndexes,['Wsets','Lsets','Winner']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wsets</th>\n",
       "      <th>Lsets</th>\n",
       "      <th>Winner</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0</td>\n",
       "      <td>Federer R.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0</td>\n",
       "      <td>Enqvist T.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0</td>\n",
       "      <td>Ferrero J.C.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3.0</td>\n",
       "      <td>0</td>\n",
       "      <td>Federer R.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3.0</td>\n",
       "      <td>0</td>\n",
       "      <td>Federer R.</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Wsets Lsets        Winner\n",
       "0   2.0     0    Federer R.\n",
       "1   2.0     0    Enqvist T.\n",
       "2   2.0     0  Ferrero J.C.\n",
       "3   3.0     0    Federer R.\n",
       "4   3.0     0    Federer R."
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Re-setting the dataframe indexes\n",
    "FedererSub.reset_index(inplace=True)\n",
    "del(FedererSub['index'])\n",
    "FedererSub = FedererSub.astype(dtype='str')\n",
    "FedererSub.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique values in Wsets feature : ['2.0' '3.0' '1.0' '0.0' 'nan']\n",
      "Unique values in Lsets feature : ['0' '1' '2' '1.0' '0.0' '2.0' 'nan' '`1']\n"
     ]
    }
   ],
   "source": [
    "print(\"Unique values in Wsets feature : \"+ str(FedererSub.loc[:,'Wsets'].unique())) \n",
    "print(\"Unique values in Lsets feature : \"+ str(FedererSub.loc[:,'Lsets'].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Define a \n",
    "c1,c2,c3 = \".\",\"nan\",\"`\"\n",
    "def toInt(s):\n",
    "    \"\"\"\n",
    "    Overview : \n",
    "    Helper function for strings cleaning + to integer transformation \n",
    "    \n",
    "    Arguments : \n",
    "    - s : String\n",
    "    \n",
    "    Returns :\n",
    "    - Integer number in String s\n",
    "    \n",
    "    \"\"\"\n",
    "    if(c1 in s):\n",
    "        s = int(float(s))\n",
    "        return(s)\n",
    "    elif(c2 in s):\n",
    "        s=0\n",
    "        return(s)\n",
    "    elif(c3 in s):\n",
    "        s = int(s.replace(c3,\"\"))\n",
    "        return(s)\n",
    "    else:\n",
    "        s = int(s)\n",
    "        return(s)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "FedererWSets = 0\n",
    "for i,val in enumerate(FedererSub['Winner']):\n",
    "    if(val==\"Federer R.\"):\n",
    "        FedererWSets += toInt(FedererSub.loc[i,'Wsets'])\n",
    "    else:\n",
    "        FedererWSets += toInt(FedererSub.loc[i,'Lsets'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2685"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FedererWSets"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So “Federer R.” won a total of **2679** Set. Great player!  \n",
    "This result makes sense since we've found, in the previous question that \"Federer R.\" won 1108 match, and to win a match you need to win 2 to 3 sets, so the result must be at least equal to 1108*2.  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "--------\n",
    "### Q3) **How many sets did the player “Federer R.” win during the years 2016 and 2017 ?**  \n",
    "1 - Filter the rows with date in (2016,2017) and exploit the previous 'Federer R.' indexes to obtain the subset.  \n",
    "2 - We do the same as before (Step 2->4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count                   52174\n",
       "unique                   4628\n",
       "top       2002-08-26 00:00:00\n",
       "freq                      127\n",
       "first     2000-01-03 00:00:00\n",
       "last      2018-10-21 00:00:00\n",
       "Name: Date, dtype: object"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Quick exploration of the 'Date' variable\n",
    "df_atp['Date'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Selecting indexes \n",
    "sub1 = lambda x: x.year in (2016,2017)\n",
    "FedIndexes1617 = df_atp['Date'].apply(sub1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Intersection with the previous indexes of Federer's observations => Federer + (2016,2017)\n",
    "FedIndexes1617 = (FedIndexes1617 & FedIndexes) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(86, 3)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Selecting feautures \n",
    "FedererSub1617 = df_atp.loc[FedIndexes1617, ['Wsets','Lsets','Winner']].copy()\n",
    "FedererSub1617.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Re-indexing data\n",
    "FedererSub1617.reset_index(inplace=True)\n",
    "del(FedererSub1617['index'])\n",
    "FedererSub1617 = FedererSub1617.astype(dtype='str')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "FedererWSets1617 = 0\n",
    "for i, val in enumerate(FedererSub1617['Winner']):\n",
    "    if (val == \"Federer R.\"):\n",
    "        FedererWSets1617 += toInt(FedererSub1617.loc[i,'Wsets'])\n",
    "    else:\n",
    "        FedererWSets1617 += toInt(FedererSub1617.loc[i,'Lsets'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "182"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FedererWSets1617"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So \"Federer R.\" won **182** set in the ATP matches of 2016 & 2017.   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "--------------------------\n",
    "### Q4)**For each match, what is the percentage of victories of the winner in the past ?**  \n",
    "If we call 'victory' during a given match a game won in a certain set of that match, then the percentage of victories of the winner in the past is expressed as the ratio :  \n",
    "\\begin{equation*}\n",
    "\\frac{W1+W2+W3+W4+W5}{W1+W2+W3+W4+W5+L1+L2+L3+L4+L5}\n",
    "\\end{equation*}\n",
    "To calculate it, we need to :  \n",
    "1 - Select 'Wi' and 'Li' columns (i in 1:5).  \n",
    "2 - Drop observations having NAs in the first Set of the match.  \n",
    "3 - Clean values in the selected features (nan, blank spaces..).  \n",
    "4 - Compute the ration for each observation.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['L1', 'L2', 'L3', 'L4', 'L5', 'W1', 'W2', 'W3', 'W4', 'W5'], dtype='object')"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Select the columns names\n",
    "names = df_atp.columns\n",
    "setColumns = names[[bool(re.search(r\"([WL][1-5])\",name)) for name in names]]\n",
    "setColumns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Subset the data with the columns names\n",
    "df_atp_sub = df_atp.loc[:,setColumns].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>L1</th>\n",
       "      <th>L2</th>\n",
       "      <th>L3</th>\n",
       "      <th>L4</th>\n",
       "      <th>L5</th>\n",
       "      <th>W1</th>\n",
       "      <th>W2</th>\n",
       "      <th>W3</th>\n",
       "      <th>W4</th>\n",
       "      <th>W5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4.0</td>\n",
       "      <td>2</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>6.0</td>\n",
       "      <td>6</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3.0</td>\n",
       "      <td>3</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>6.0</td>\n",
       "      <td>6</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7.0</td>\n",
       "      <td>5</td>\n",
       "      <td>3</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>6.0</td>\n",
       "      <td>7</td>\n",
       "      <td>6</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.0</td>\n",
       "      <td>4</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>6.0</td>\n",
       "      <td>6</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6.0</td>\n",
       "      <td>7</td>\n",
       "      <td>4</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>7.0</td>\n",
       "      <td>5</td>\n",
       "      <td>6</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    L1 L2   L3  L4  L5   W1 W2   W3  W4  W5\n",
       "0  4.0  2  NaN NaN NaN  6.0  6  NaN NaN NaN\n",
       "1  3.0  3  NaN NaN NaN  6.0  6  NaN NaN NaN\n",
       "2  7.0  5    3 NaN NaN  6.0  7    6 NaN NaN\n",
       "3  1.0  4  NaN NaN NaN  6.0  6  NaN NaN NaN\n",
       "4  6.0  7    4 NaN NaN  7.0  5    6 NaN NaN"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_sub.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Find indexes with missing values in the first set (W1 or L1)\n",
    "naIndexes = df_atp_sub['W1'].isna() | df_atp_sub['L1'].isna()\n",
    "cleanIndexes = naIndexes.apply(lambda e : not e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Filter data with cleanIndexes\n",
    "df_atp_sub = df_atp_sub.loc[cleanIndexes,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([6.0, 7.0, 3.0, 4.0, 2.0, 5.0, 0.0, 1.0, nan, ' ', 10.0, 9.0, 8.0,\n",
       "       13.0, 12.0, 20.0, 11.0, 21.0, 16.0, 15.0, 14.0, 17.0, 70.0, 18.0,\n",
       "       19.0, 22.0, 26.0], dtype=object)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Unique values in Wi features (i in 1:5)\n",
    "pd.concat([df_atp_sub.W1,df_atp_sub.W2, df_atp_sub.W3, df_atp_sub.W4, df_atp_sub.W5]).unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([4.0, 3.0, 7.0, 1.0, 6.0, 2.0, 0.0, 5.0, nan, ' ', 8.0, 11.0, 10.0,\n",
       "       18.0, 9.0, 19.0, 14.0, 13.0, 12.0, 15.0, 68.0, 16.0, 17.0, 20.0,\n",
       "       24.0], dtype=object)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Unique values in Li features (i in 1:5)\n",
    "pd.concat([df_atp_sub.L1,df_atp_sub.L2, df_atp_sub.L3, df_atp_sub.L4, df_atp_sub.L5]).unique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since the values are not all numerical, we'll need to deal with the special cases and avoid including nan values in the summation.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function : handles missing values/special cases + calculates the winning percentage\n",
    "def WinPercentage(obs):\n",
    "    myobs = obs.copy()\n",
    "    myobs[myobs == ' '] = np.nan\n",
    "    wins = np.nansum(myobs['W1':'W5'])\n",
    "    loses = np.nansum(myobs['L1':'L5'])\n",
    "    if wins+loses == 0:\n",
    "        return(0)\n",
    "    else:\n",
    "        per = 100*wins/(wins+loses)\n",
    "        return(per)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Execution will take about 2 minutes\n",
    "df_win_per = df_atp_sub.apply(axis=1, func=WinPercentage)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wper</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>51914.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>61.913378</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>8.635817</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>56.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>60.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>66.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               Wper\n",
       "count  51914.000000\n",
       "mean      61.913378\n",
       "std        8.635817\n",
       "min        0.000000\n",
       "25%       56.000000\n",
       "50%       60.000000\n",
       "75%       66.666667\n",
       "max      100.000000"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_win_per = pd.DataFrame({'Wper':df_win_per})\n",
    "df_win_per.describe(include='all')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The best way to represent these results is to use a density plot (or a histogramm) :   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (283467994)>"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pln.ggplot(data=df_win_per, mapping=pln.aes(x='Wper'))+pln.geom_density(colour=\"brown\", fill=\"orange\")+pln.theme_bw()+pln.xlab(\"\")+pln.ylab(\"Density\")+pln.xlab(\"Percentage of winner's victories\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can have a \"look from above\" of our density plot using a boxplot :  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjkAAAGcCAYAAAA70rGtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAAPYQAAD2EBqD+naQAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XtclHX+///nDINyUkEFQQVSV8tjqW27ns1sK7ODZn78pZWWh7JSd+28uopph7WDZrRam261mmbZSTvsipbamlq5paVZhkdAEI8goMNcvz9c5isyIBfMMMM1j/vt5i3meg8zz+nlJS/e13W9L5thGIYAAAAsxu7vAAAAAL5AkwMAACyJJgcAAFgSTQ4AALAkmhwAAGBJNDkAAMCSaHIAAIAl0eQAAABLoskBAACWRJMDAAAsiSYHAABYksPfAfzh2LFjOnXqlL9jAAAAEyIiIhQdHV3p5wddk3Ps2DGlpqbqzJkzHsftdrs6d+6srVu3yuVy1XC6mhcVFeX+vHl5ef6O43PU19qor7VRX2urTH1DQ0N13333VbrRsQXbXcgzMjL0yiuvaPDgwWrcuLG/4/idw+FQTEyMjh49KqfT6e848DLqa23U19qob2mHDx/WihUrNHbsWDVt2rRS3xN0MzklGjdu7PF/ksvlUlZWluLj42W3W/+UJcMw5HQ6FRsbK5vN5u84Pkd9rY36Whv1tTZf1Nf6f0sAAEBQoskBAACWRJMDAAAsiSYHAABYEk0OAACwJJocAABgSTQ5AADAkmhyAACAJfl9McCVK1dqzZo12rNnj7p166aHHnrIPbZ3717NmzdPe/bsUXx8vO699161b9++1Pe+8847KigoUNeuXXX//fcrIiLCHx8DAAAEGL/P5DRs2FBDhw7VH/7wh1LbnU6nZs6cqW7duumtt97SLbfcolmzZrnv37F161YtXbpUf/nLX7Ro0SKdOXNGCxYs8MdHAAAAAcjvTU737t31+9//XvXr1y+1fdu2bSoqKtKgQYMUGhqqK6+8Uk2aNNF//vMfSdKaNWt01VVXqWXLloqIiNDw4cO1YcMGFRUV+eNjAPCjJffeqyXt22tpx476/OqrtbRjR39HAhAA/N7klGffvn1KTk4udf+KFi1aaN++fZLOHspq0aKFeyw5OVkul0sZGRk1nhWAn61bV2bTknMObQMITn4/J6c8BQUFioyMLLUtMjJSp06dkiQVFhaWGrfZbIqIiFBBQUGZ18rMzFRmZqYkKScnR/n5+ZLk8VbuJdvKu8271RiGIZfLJZfLFTQ3gDv3v1YXDPWtaNZmybhxGva3v9VgmpoVDPU9F/uvtfmivgHb5ISHh7sbmhKnTp1SeHi4JCksLKzC8XMtWLBAKSkp7sfDhg2TJGVlZZX7/tnZ2VXOjsBHfYPEhg0V7ueondh/rc2b9Q3YJicpKUkrVqyQy+VyH7JKT0/XtddeK+ns4an09HT16dNH0tnDV3a7XU2bNi3zWuPGjdONN94o6exMzurVqyVJ8fHxZZ7rcrmUnZ2tuLg4r93qPZAZhiGn0ymHwxE0vylQX4vp2VPasMHj0LBt22o4TM0Kivqeg/3X2i5U36r8wuL3Jqe4uFjFxcXuKbnTp0/LbrerY8eOCg0N1fvvv68bbrhB//nPf5SVlaVu3bpJkvr166fnn39effr0UZMmTbR48WL17NlTdevWLfMeCQkJSkhIkCRlZGRo48aNklThTmK324NmJyr5rMGwE5WgvtZx24IF5Z5/Y/UaB0N9PWH/tTZv1tfvTc6yZcu0dOlS9+Mvv/xS/fr106RJkzRlyhS99NJLWrJkiZo0aaLHH39c9erVkyR17txZ//d//6eUlBSdOnVKXbt21bhx4/z1MQD40W0//FCm0bnthx/8lAZAoPB7k3Pbbbfptttu8zh20UUX6dlnny33ewcOHKiBAwf6KhqAWuS2H36Qy+VSVlaWx0PRAIKP9ef7AABAUKLJAQAAlkSTAwAALIkmBwAAWBJNDgAAsCSaHAAAYEk0OQAAwJJocgAAgCXR5AAAAEuiyQFgGSf37VMedx0H8D9+v60DAFTXvtWrtWHiRPfjryX1WbBAzXr29F8oAH7HTA6AWu/cBqfEF9ywFwh6NDkAarXD27eXO5aXkVGDSQAEGpocALWay+n0dwQAAYomB0CtFnfZZeWORTVtWoNJAAQamhwAtV6nSZPKbOv65z/7IQmAQEKTA6BWMwxDPy9bVmb7riVL/JAGQCChyQFQq53Jy1NBZmaZ7SfT01VcVOSHRAACBU0OgFrNqGgsJKTGcgAIPDQ5AGq1UxVcJl6Um1uDSQAEGpocALXaJ4MHlzv2Qb9+NZgEQKChyQEAAJZEkwOgVhv2/ffljv1/FayGDMD6aHIA1GrO/Pxyx7i6CghuNDkAarXQevXKHXOEhdVgEgCBhiYHQK124PPPyx07vGNHzQUBEHBocgDUauHx8eWORTRuXINJAAQamhwAtVrjtm3LHYuIja3BJAACDU0OgFptxbXXljv2+YMP1mASAIGGJgdArRYeF1fuWGRiYg0mARBoaHIA1GrXvfFGuWO/nTixBpMACDQ0OQBqtZMnT/o7AoAARZMDoFb7cc6ccsd+euedGkwCINDQ5ACo1S4eOrTcsabXXFODSQAEGpocALVa9MUXlztWr4LVkAFYH00OAACwJIe/A/hDVFSUHA6HDMMoM2YYhnvM07jVlHzGYPisEvW1oqMV3Gn82LFjatCgQQ2mqVnBUN9zsf9a24Xq63CYb1mCssnp3LmzYmJi5HQ6PY7HxMTI5XLJ5XLVcDL/KS4u9neEGkN9rWXb66+XO7Z/xQpF3n57DabxDyvX93zsv9ZWUX1jYmJMv15QNjlbt25Vx44dFethyXeXy6Xc3Fw1atRIdrv1j+YZhqHi4mKFhITIZrP5O47PUV/ruXTKFB385BOPYx1GjarhNDUrGOp7LvZfa7tQfXNycky/pukmZ+/evTpx4oQ6duwoSSoqKtKzzz6rHTt2qH///ho5cqTpEDUtLy9PTqfT418am83mHguGv1QlguXzUl/rqRcRUe6Y4XLJHhJSg2n8w8r1PRf7r7VdqL7lHX2piOlWeMyYMXrzzTfdjx955BGlpKRo586dGjt2rF5++WXTIQCgqlxnzpQ7ZgTRND+Askw3Od9995169eol6WxX9frrr+uZZ57R119/renTp+tvf/ub10MCQHlCIyPLHQupU6cGkwAINKabnBMnTrivVti0aZNOnDihYcOGSZJ69uypX3/91bsJAaACS9q3r9IYAOsz3eQ0b95cX331lSRpxYoVateunRISEiRJR48eVUQFx8cBAABqiukTj++++25NmTJFy5cv19atW/XCCy+4x7766iu1bdvWqwEBoKpu++EHf0cA4Eemm5xHH31UTZs21ZYtWzR+/PhSV1MdPXpUo0eP9mY+AKjQbT/8wGEpAB5VaZ2cO+64Q3fccUeZ7fPnz692IAAwI//QoXLHTp84oTr169dgGgCBpMqrKX366ad64oknNHbsWO3bt0+StG7dOmVkZHgtHABciBFEK98CMMf0TE5OTo5uuukmbdq0SQkJCcrMzNQ999yjpKQkLVy4UJGRkUpNTfVFVgAoI+p/Fz6czx4ayiwOEORMz+RMmjRJhw8f1rZt27Rnz55SN9Hq37+/0tLSvBoQAKqiokUCAQQH003OqlWrNGvWLLVr167MssuJiYk6cOCA18IBwIXsfOutcsd+/vnnGkwCINCYbnKcTqciy1lh9OjRo6rDCqMAatC3c+aUO7Zr2rQaTAIg0Jhucn73u99p4cKFHseWLl2qHj16VDsUAFTWb//5z3LHrl+ypAaTAAg0pk88njlzpq688kr17t1bQ4YMkc1m0/vvv6+nnnpKq1at0oYNG3yREwA8at26tbb4OwSAgGR6Jqdbt25au3atbDabJk+eLMMwNGvWLGVmZiotLU1dunTxRU4AKFeZlY3tdlY7BlC1xQC7deumL774QgUFBTp69Kiio6O5ZxUAv7rthx/kcrmUlZWl+Ph4f8cBEACqvBigJIWHh6tp06Y0OAACwumTJ5X+xhs6ffKkv6MACACVmsmZMGGCHnzwQSUlJWnChAkVPtdms2nu3LleCQcAZpzJy9PeN99Up9tvV1iDBv6OA8DPKtXkfPTRR7r77ruVlJSkDz/8sMz6OOeiyQEAAIGgUk1Oenq6++s9e/b4KgsAAIDXmDonp7CwUA888IA2b97sqzwAAABeYarJCQsL06JFi1RQUOCrPAAAAF5h+uqq7t27a9OmTb7IAgAA4DWm18mZMWOGRowYIYfDoeuuu05xcXFlTkRu2LCh1wICAABUhekmp3v37pKkBx98UA899JDH5xQXF1cvFQAAQDWZbnIWLlxY4SXkAAAAgcB0kzNy5EgfxAAAAPCuKt27Sjq7Xs6XX36pI0eOqGHDhurZs6eSk5O9mQ0AAKDKTDc5xcXFuueee7Ro0SK5XC73drvdrrvuukvz58+X3V6tW2IBAABUm+luZPr06XrjjTc0c+ZMpaenq6CgQOnp6Zo1a5beeOMNpaSk+CInAACAKaZnckoamUcffdS9LTk5WY888ogMw9Df/vY3Gh0AAOB3pmdyDh06pK5du3oc69q1qw4dOlTtUAAAANVluslp2bKlPvroI49jH330kVq2bFntUAAAANVl+nDVpEmTdM899ygnJ0dDhw5VfHy8Dh06pLffflvLli3T/PnzfZETAADAFNNNztixY1VUVKQnnnhCy5Ytk81mk2EYio2N1YsvvqgxY8Z4LdzQoUNLPT59+rQuv/xyTZkyRZI0evRoHTt2zH01V2xsrFJTU732/gAAoPaq0jo5DzzwgO677z7t3LlTR48eVcOGDXXxxRd7/dLxt99+2/11cXGx7r77bvXo0aPUcx577LFyzxECAADBy3RXMmPGDGVkZMhut6tdu3bq0aOH2rZtK7vdrszMTM2YMcMXOfXtt9+qsLDQfe8sAACAiphuclJSUnTgwAGPYxkZGT67fDwtLU29evVS3bp1S22fM2eORowYoccff1w//vijT94bAADUPqYPVxmGUe4NOjMzMxUdHV3tUOc7ceKENm/erKeeeqrU9j/96U9q1aqVpLNNUEpKiubNm6e4uLgyuTIzMyVJOTk5ys/Pl6RSKzaXKNnmacyKDMOQy+WSy+UKihuvUl9rO7e+wVDjYK5vMKC+1VepJuett97SW2+9JUmy2WyaPHlymWamsLBQX3/9dZlzZrzh888/V0JCgi6++OJS29u1a+f+esCAAVq/fr2++eYbXXfddaWet2DBglIzTMOGDZMkZWVllfue2dnZ3oiOAEV9rakwN1eSlJubq/yQED+nga+w/1qbN+tbqSbn9OnTOnnypKSznWV+fr5CzvsHpE6dOrrjjjv08MMPey1cibS0NPXv3/+Cz7Pb7TIMo8z2cePG6cYbb5R0diZn9erVkqT4+Pgyz3W5XMrOzlZcXFxQ3IPLMAw5nU45HI6g+U2B+lrXyeJiSVKjRo1Uz8P+bTXBVl/2X2u7UH0rmpgoT6WanDvvvFN33nmnJOnKK6/Uyy+/rLZt25p+s6rYvXu39u3bp759+5banpOTo+zsbLVp00aStGbNGv3888+6//77y7xGQkKCEhISJJ09b2jjxo2SVOFOYrfbg2YnKvmswbATlaC+1lRSU+prbdTX2rxZX9Pn5Kxdu9Yrb1xZq1ev1uWXX66YmJhS2wsKCvTKK68oMzNTDodDiYmJmjp1qruZAQAAwc10k/PnP/9Zhw8f1oIFC8qMjRs3Tk2aNPHqZeTjxo3zuD0pKUlz58712vsAAABrMT0f9NZbb6lnz54ex3r16uU+QRkAAMCfTDc5GRkZSkxM9DjWvHnzctfQAQAAqEmmm5zY2Fht377d49j27dvVsGHDaocCAACoLtNNzs0336zp06dr8+bNpbZv3rxZM2bM0KBBg7wWDgAAoKpMn3g8c+ZMffnll+rWrZvatm2rpk2bKiMjQzt27NBll12mWbNm+SInAACAKaZncho0aKCvvvpK8+fPV8eOHSVJHTt21CuvvKKNGzeqQYMGXg8JAABglumZHOns6sZjxozRmDFjvJ0HAADAK6y/ZCQAAAhKlZrJqV+/vtauXauuXbuqXr16FS4vbbPZdPz4ca8FBAAAqIpKNTmTJ0923y5h8uTJQXUPDQAAUDtVqsmZNm2a++vp06f7KgsAAIDXmD4nZ/78+Tpy5IgvsgAAAHiN6SZn4sSJSkhI0IABA7R48WLl5eX5IhcAAEC1mG5ysrKy9NJLL6moqEh33nmnmjRpoqFDh+qDDz7Q6dOnfZERAADANNNNTkxMjMaMGaO0tDQdPHhQTz75pA4cOKBBgwapSZMmuvvuu32REwAAwJRqrZPTpEkTTZw4Uf/5z3/06aefKjIyUv/4xz+8FA0AAKDqqrTicYkDBw5o6dKlWrp0qbZu3arGjRvr3nvv9VY2AACAKjPd5OTk5Gj58uV66623tHHjRkVFRenmm2/WrFmz1L9/f4WEhPgiJwAAgCmmm5ymTZsqNDRUAwYM0LJlyzRw4EDVrVvXF9kAAACqzHST8/e//12DBw9WvXr1fJEHAADAK0w3OXfeeacvcgAAAHgVdyEHAACWVK2rq1BW4dGjKi4o8HeMSjMkFTudCnE4FAy3XXUZhgpzcpRvGLIHwY1mg62+p7Kz3f+lvtbD/ls7hISHKywmxt8xJNHkeFXh0aN6r3cvGS7D31GAoJY2YoS/IwBBy2a3adC69QHR6NDkeFFxQYEMl6Gr70xXRL0z/o4DAECNOnUyVP9+vcXZIxpWaXL27NmjX375RV26dFHDhg298ZK1WkS9M4ps4PR3DAAAgprpE48nT56sSZMmuR+/9957uvjii/WHP/xBrVu31jfffOPVgAAAAFVhusl57733dPnll7sfP/744xowYIC+//57XXHFFZoyZYpXAwIAAFSF6SYnMzNTSUlJkqTdu3frp59+0pQpU9ShQwdNmDBBX3/9tddDAgAAmGW6yWnQoIGy/3eZ5r///W81bNhQXbt2lSTVqVNHBbXo8mkAAGBdpk887t27t/7yl7/o0KFDevbZZ3XzzTe7x3766Sf3LA8AAIA/mZ7JeeGFFxQfH69HH31USUlJmjVrlnvszTffVK9evbwaEAAAoCpMz+Q0a9ZMa9as8Tj22WefKSwsrNqhAAAAqsuriwHWr1/fmy8HAABQZZVqcu66665SjxcuXOiTMAAAAN5SqSbHMLgXEwAAqF0q1eQsWrTI1zlqVFRUlBwOh8fmzTAM95jZ5o5WEACAsz8PTf8MvcDPX4fD/Bk2pr6jsLBQV1xxhWbPnq1rrrnG9JsFis6dOysmJkZOp+f7S8XExMjlcsnlcpl63eJyXg8AgGBS7HSW+zO2IhX9/I2pwg0/TTU5YWFhysjIUEhIiOk3CiRbt25Vx44dFRsbW2bM5XIpNzdXjRo1kt1u7gr7kCp0mQAAWE2Iw2F65uVCP39zcnJM5zC9Ts7gwYP1zjvvmH6jQJKXlyen0ymbzebxT0VjFf7x9wcDACAA2KQq/Ry90M9ms0xPPfTo0UOPP/64MjIydN111ykuLk42W+kf74MHDzYdBAAAwJtMNzmjRo2SJK1cuVIrV64sM26z2VRcXFz9ZAAAANVguslJT0/3RQ4AAACvMt3kJCcn+yIHAACAV1X5cqBPP/1UW7Zs0f79+zVlyhQlJSVp3bp1+s1vfqOmTZt6M2Otk5sZplMnOGQHAAguBfmBdfW16SYnJydHN910kzZt2qSEhARlZmbqnnvuUVJSkhYuXKjIyEilpqb6ImutseHdJH9HAAAg6Jm+hHzSpEk6fPiwtm3bpj179pRalbB///5KS0vzakAAAICqMD2Ts2rVKr366qtq165dmauoEhMTdeDAAa+Fq6163rJP4ZEcrgIABJeC/JCAOpphuslxOp2KjIz0OHb06FHVqVOn2qFqu0YJhYpswC0eAADBJf94YK38b/pw1e9+9zstXLjQ49jSpUvVo0ePaocCAACoLtMt18yZM3XllVeqd+/eGjJkiGw2m95//3099dRTWrVqlTZs2OCLnAAAAKaYnsnp1q2b1q5dK5vNpsmTJ8swDM2aNUuZmZlKS0tTly5dfJETAADAlCodPOvWrZu++OILFRQU6OjRo4qOjlZERIS3swEAAFRZtc4QCg8PV3h4uLeyAAAAeE2Vmpx//etfeuedd3TgwAEVFhaWGrPZbKyVAwAA/M50kzN79mw98sgjSkxM1CWXXFLu5eQAAAD+ZLrJSU1N1fjx4zVv3jzZbDZfZAIAAKg201dXHTlyRIMHD6bBAQAAAc30TM4NN9ygDRs2qF+/fr7IYwmnTob6OwIAADUu0H7+mW5yRo4cqfHjx6ugoEBXX321oqOjyzwnWNfKCQkPl81u079fb+HvKAAA+IXNblNIgFx5bbrJueaaayRJzzzzjJ555plSh60Mw5DNZitz485gERYTo0Hr1qu4oMDfUSrNkFTsdCrE4VAwHIB0GYZycnIUGxsrexAccg22+uYdOqS0ESN01T//qagmTfwdx+eCrb7sv7VDSHi4wmJi/B1DUhWanLVr1/oih2WExcRIAVLcyjAMQ06nUw6HIyjOs3K5XDppsykyPl52u+lT0mqdoKuvYUiSIuLiFNm0qZ/T+F7Q1Zf9FyaZbnL69OnjixwAAABeVa0Vj7Ozs8ssBihJSUlJ1XlZAACAajPd5OTm5uqBBx7QihUrdObMmVJjwX5ODgAACBymm5zRo0fr888/10MPPaR27dqpTp06vsgFAABQLVU68fjFF1/UHXfc4Ys8AAAAXmH69PTo6Gg1btzYF1kAAAC8xnST89BDD2nevHlyOp2+yAMAAOAVpg9X7dy5Uz/++KNatWqlPn36lFnx2Gazae7cuV4LCAAAUBWmm5yVK1e6F2Fav359mXGaHAAAEAhMNznp6em+yAEAAOBV1l8XGwAABKVKzeSsWLFC/fr1U3R0tFasWHHB5w8ePLjawQAAAKqjUk3OkCFD9NVXX+mKK67QkCFDKnyut1c8njNnjtatWyeH4/9FTU1NVWxsrCQpJydH8+bN044dOxQTE6ORI0eqe/fuXnt/AABQO1WqyUlPT1dCQoL765p200036c477/Q49uyzz6ply5b685//rB9//FFPP/20kpOT1axZsxpOCQAAAkmlmpzk5GSPX/tbRkaGfv75Z02dOlV169ZV586dddlll2nt2rUaMWKEv+MBAAA/Mn3i8dixY7V48WLt27fPF3k8+uyzz3TbbbdpwoQJ+ve//+3evnfvXsXGxioqKsq9rUWLFjWaDQAABCbTl5D/+OOPeuONN3TmzBklJiaqT58+6t27t3r37q3WrVt7PeANN9ygu+66S5GRke7DUZGRkerevbsKCwsVGRlZ6vlRUVEqKCgotS0zM1OZmZmSzp7Dk5+fL0lyuVxl3q9km6cxKzIMQy6XSy6XSzabzd9xfI76Wtu59Q2GGgdzfYMB9a0+003Ohg0bVFRUpI0bN2rdunVav369Jk2apFOnTqlJkybq3bu3li5d6rWArVq1cn/dsWNHXX/99fryyy/VvXt3hYWF6dSpU6Wen5+fr/Dw8FLbFixYoJSUFPfjYcOGSZKysrLKfd/s7GxvxEeAor7WVJibK0nKzc1VfkiIn9PAV9h/rc2b9TXd5EhS3bp11bdvX/Xt21f5+flau3atnnvuOX3xxRdavny5V5uc89lsNhmGIens+UHZ2dnKz893z+ikp6crKSmp1PeMGzdON954o6SzMzmrV6+WJMXHx5d5fZfLpezsbMXFxblXdrYywzDkdDrlcDiC5jcF6mtdJ/93ZWejRo1Uz8P+bTXBVl/2X2u7UH0rmpgoj+km5/jx49qwYYO++OILffHFF9q6dasaNGignj176rnnnlOfPn1Mh6jIhg0b1KVLF4WFhWnnzp1atWqVxo4dK0lq2rSpfvOb32jJkiW64447tHPnTm3dulW33357qddISEhwXx2WkZGhjRs3SlKFO4ndbg+anajkswbDTlSC+lpTSU2pr7VRX2vzZn1NNzmNGjVS3bp1dcMNN2jkyJF67bXX1KFDB6+E8WTlypVKTU2Vy+VS48aNNXz4cPXu3ds9/tBDD+nFF1/U8OHDFRMTo4kTJ3L5OAAAMN/kdOjQQdu3b9e6deskne00bTab2rdv7/VwkvT0009XOB4bG6snnnjCJ+8NAABqL9PzQf/973915MgRvfrqq0pOTtabb76pLl26KDY2VoMGDdKcOXN8kRMAAMCUKh30ql+/vq6//no988wzWrdunT766CN16NBBH3zwgSZPnuztjAAAAKaZPlxVVFSkr776yn3i8aZNm1RQUKAmTZro1ltv9fqJxwAAAFVhuslp0KCBeyHA3r17a+7cuT5bCBAAAKCqTDc5f//739WnTx8lJib6Ig8AAIBXmG5yuPElAACoDay/mhIAAAhKNDkAAMCSaHIAAIAlVarJ+fDDD3X8+HFfZwEAAPCaSjU5gwYN0k8//SRJCgkJ0ebNm30aCgAAoLoq1eQ0aNBAOTk5kv7fvaoAAAACWaUuIe/Tp49uv/12derUSZJ07733qn79+h6fa7PZlJaW5r2EAAAAVVCpJmfhwoX661//qp07d8pmsyk8PFyRkZG+zgYAAFBllWpyYmJi9NRTT0mS7Ha7nnvuOV1xxRU+DQYAAFAdplc8drlcvsgBAADgVaabHEnKzc1Vamqq1q9fryNHjqhhw4bq3bu3xo8fr0aNGnk7IwAAgGmmFwPcvXu3OnbsqFmzZsnpdKpNmzZyOp2aOXOmOnXqpN27d/siJwAAgCmmZ3ImT56s6OhoffXVV0pKSnJv379/v6699lo9+OCDeu+997waEgAAwCzTMzlr167VjBkzSjU4kpSYmKjp06drzZo1XgsHAABQVaabHJfLJYfD8wSQw+HgxGQAABAQTDc5PXr00BNPPKEjR46U2n706FHNmjVLPXv29Fo4AACAqjJ9Ts5zzz2nXr16KTk5Wf369VP2Sm6UAAAdSklEQVR8fLwOHTqktLQ01alTR6+//rovcgIAAJhieianffv2+v777zVmzBhlZmZqzZo1yszM1NixY/Xdd9+pffv2vsgJAABgSpXWyWnevLmef/55b2cBAADwGtMzOQAAALUBTQ4AALAkmhwAAGBJNDkALMF56pQy16+XJBUXFfk5DYBAQJMDoNbL2bpVy7t109dPPCFJ+vjmm5W7bZufUwHwtyo3OTt27NCbb76pJ598UllZWZKkX375RSdPnvRaOACojLRRo2Q4nf9vQ3GxVo8c6bc8AAKD6UvIT506pdGjR+vtt9+WJBmGoWuvvVbx8fF67LHH1KJFC/31r3/1elAA8CT/0CG5zpwps724sFCn8/JUJyrKD6kABALTMzkPPvig1qxZo5UrV+r48eMyDMM9NmDAAH366adeDQgAFbGd82/Q+QzupQcENdMzOe+8845mz56ta6+9VsXFxaXGLrroIu3Zs8db2QCgWmzl3EwYQHAwPZOTl5enhIQEj2P5+fnVDgQAZrjOPRfn/LHTp2swCYBAY7rJ6dSpk959912PY6tWrdLll19e7VAAUFlRzZtLNlvZAZtNYdHRNR8IQMAwPZc7depU3XTTTTp16pRuvfVW2Ww2bd68WW+99ZYWLlyojz/+2Bc5vSoqKkoOh6PU+UQlDMNwj3kat5qSzxgMn1WivlbkKi6WzW6Xcd7hc3toqFwul2yeGiCLCIb6nov919ouVF9HFQ4/m/6O66+/XkuXLtVDDz2kxYsXS5LGjx+v5s2ba/HixbrqqqtMh6hpnTt3VkxMjJzlTHPHxMTI5XLJFUQnLZ5/fpWVUV9rOX3iRJkGRzp7qOp0fr5CwsL8kKpmWbm+52P/tbaK6hsTE2P69ap0Vt6QIUM0ZMgQ7dq1S4cPH1bDhg11ySWXVOWl/GLr1q3q2LGjYmNjy4y5XC7l5uaqUaNGstutv1aiYRgqLi5WSEiIpX/jLUF9rcceHX32cNX5v/nZ7apr8cvHg6G+52L/tbYL1TcnJ8f0a1br0oM2bdqoTZs21XkJv8jLy5PT6fT4l8Zms7nHguEvVYlg+bzU13qK8/LKNjiS5HKpuKhIjiCYybFyfc/F/mttF6pveUdfKmK6yZkxY0a5Y3a7XQ0aNNBll12mXr16mQ4DAGbZK2hiuIQcCG6m/wWYPXu2nE6niv53A7zQ0FCd+d9qo3Xr1pXT6ZTL5VKXLl308ccfezwkBADeYlT0210QnbcBoCzTBzU///xzNW/eXIsWLdKRI0dUVFSkI0eO6LXXXlPz5s31xRdf6LPPPtOBAwf00EMP+SIzALgZFZ2bERJSc0EABBzTMzn33XefHnzwQd15553ubdHR0Ro1apQKCwv1xz/+UZs2bdKUKVMqPLQFAN6Qv29fuWOFhw8rskmTGkwDIJCYnsn57rvvlJyc7HGsRYsW2rZtmySpQ4cOOn78ePXSAcAFhFXQxLAYIBDcTDc5ycnJ+vvf/+5x7JVXXnE3QLm5uWrcuHH10gHABdSNivK44rEtJET2OnX8kAhAoDB9uOqpp57S0KFDdfHFF2vgwIGKjY1VTk6OVq5cqV9//VXLly+XJKWlpalPnz5eDwwA57I7HGo/dqx+fPVV913HbXa7Oo4fHxSX3QIon+kmZ9CgQdq8ebOeeuopvffee8rMzFRCQoJ++9vfatmyZbrsssskSampqV4PCwCedHrgAYU1aqT0Dz7QGadTFw8dqtb/93/+jgXAz6q0iETnzp319ttvezsLAFTJsWPH9M2TT7off/3EE2ozbJgfEwEIBNZfFxuA5X3Ss2eZbUvat/dDEgCBpEozOb/88ov+8Y9/aNeuXSosLCwz/uGHH1Y7GABUxuGtW8sdO3r0aJVu6gfAGkw3OVu2bFGfPn2UnJysXbt2qVOnTjp+/Lj27Nmj5s2b6ze/+Y0vcgKAR9sWLix3bN+yZYq5554aTAMgkJg+XPXwww/r1ltv1fbt22UYhl577TX9+uuv2rBhg+x2ux555BFf5AQAjy6rYNHRS2lwgKBWpcUAb7vtNvdt0EsOV3Xv3l3Tpk3To48+6t2EAFCBmJgYj+vkcEsHAKabHJvNpjp16shmsykuLk579+51jzVv3ly7du3yakAAuJDbtm9XvXMOlTdo21a3ff+9HxMBCASmz8lp166ddu/erSuvvFLdunXTc889p44dOyo0NFRPP/20WrVq5YucAFChGz74QC6XS1lZWYqPj/d3HAABwHSTM3bsWPfszZNPPqk//OEPuvTSSyVJkZGReuedd7ybEAAAoApMNzm33367++u2bdtqx44d2rhxowoKCvT73/9ecXFxXg0IAABQFabPyXnjjTeUm5vrfhwVFaWrr75aN954oxwOh9544w2vBgQAAKgK003OqFGjtHv3bo9j6enpGjVqVLVDAYBZR3bs0Pa//U173nxTx7gAAoCqcLjKMIxyx44ePap69epVKxAAmHVgzRqtnzhRNodDhmFo75Il6vPSS2raq5e/owHwo0o1OZ988ok++eQT9+PnnntOTZo0KfWcwsJCrVmzxn0XcgCoCYZhaONjj8lwuWScPu3evvGxx3TLhg1+TAbA3yrV5OzatUsfffSRpLPr5Kxfv15169Yt9Zw6deqoQ4cOevKcOwEDgK+dOXFCZ/LyymwvOnpUzsJCOcLC/JAKQCCoVJMzceJETZw4UZLUokULvf/+++7LxgHAn0Lr1ZMjPFzOgoJS2+vUr6+Q834ZAxBcTJ94nJ6eToMDIGDY7Hb9dvr0Mtt/98QTsnm63QOAoGH6xGPp7AnGn3zyiQ4cOOC+d1UJm82mqVOneiUcAFTGnpUry2xL//BDJfbv74c0AAKF6SbnX//6l4YMGaK8vDzVqVNHoaGhpcZpcgDUJOfp08pcv77M9gNpaXK5XO6bCQMIPqb3/smTJ6tr1676+eefVVhYqJMnT5b6c+LECV/kBACPCnNyyh07zb9HQFAzPZPz66+/6vnnn+dGnAACQkRCgmS3Sy5Xqe22kBCFRUf7KRWAQGB6JqdLly7av3+/L7IAgGl2u13tx44ts73jAw/4IQ2AQGK6yXn55Zf14osv6rPPPpPT6fRFJgCoNMMw9Ot775XZ/uu77/ohDYBAYvpwVffu3XXmzBkNGDBAdrtd4eHhpcZtNpuOHz/utYAAUJEzJ0+q4NChMtvz9u9nMUAgyJluciZPnszaEwAChiMi4uw9q86bWQ6pW5fFAIEgZ7rJme5h0S0A8Be7w6GO992nbfPmyfjfycc2u12dJkzgFzIgyFVpMcAS+/fv1/79+3XppZcqMjLSW5nczpw5o/nz5+u7777TyZMn1bhxY916663q27evJGn06NE6duyYex2M2NhYpaamej0HgMDWfswYhcfGas/KlTp9+rQuHjJELW+6yd+xAPhZlZqcV155RSkpKcrMzJTNZtOWLVvUpUsXDRo0SH379nXf56q6iouL1bBhQ82cOVNxcXHauXOnZsyYofj4eF1yySWSpMcee0xdu3b1yvsBqJ1sNptaDRqkFjfdpKysLMXHx/s7EoAAYPrqqjlz5uiBBx7QiBEj9Mknn8gwDPdY37599fbbb3stXFhYmIYPH674+HjZ7Xa1a9dObdu21Y4dO7z2HgAAwJpMz+TMmzdPU6dO1ZQpU1RcXFxq7OKLL9ZPP/3ktXDnKyws1C+//KIbbrjBvW3OnDkyDENJSUkaMWKE2rVr57P3BwAAtYfpJufgwYPq3r27x7HQ0FDl5+dXO5QnhmFo7ty5at26tTp37ixJ+tOf/uReeTktLU0pKSmaN2+e4uLiSn1vZmamMjMzJUk5OTnujK7zVkg9d5unMSsyDEMul0sulysoTtKkvtZGfa2N+lqbL+pruslJTk7W5s2b1a9fvzJjmzZtUps2bbwS7FyGYejll19Wbm6uZsyY4S72ubM2AwYM0Pr16/XNN9/ouuuuK/X9CxYsUEpKivvxsGHDJElZWVnlvmd2drY3PwICDPW1NuprbdTX2rxZX9NNzpgxYzR9+nTFxsZq8ODBks5eBbVq1SrNnj1bs2bN8lo46WyDM3/+fP3666964oknFFbBwl52u73UOUIlxo0bpxtvvFHS2Zmc1atXS5LHkxNdLpeys7MVFxcXFHcvNgxDTqdTDocjaH5ToL7WRX2tjfpa24XqW9HERHlMNzkPPvig9u3bp7Fjx2rcuHGSpB49ekiSxo8fr/Hjx5sOUZEFCxbop59+0syZMxUREeHenpOTo+zsbPfM0Zo1a/Tzzz/r/vvvL/MaCQkJSkhIkCRlZGRo48aNklThTmK324NmJyr5rMGwE5WgvtZGfa2N+lqbN+tbpUvIX3zxRU2aNEmrV6/W4cOH1bBhQ1111VVq3bq1V0KVyM7O1scff6zQ0FDddddd7u1DhgzR73//e73yyivKzMyUw+FQYmKipk6d6m5mAABAcKvyYoAtW7bUWA93/vWmuLg4ffjhh+WOz50716fvDwAAai/T80HLli3T7NmzPY49++yzWr58ebVDAQAAVJfpJufpp59W3XJuehceHq6nn3662qEAwKxT2dlK/+gjZa1erYLDh/0dB0AAMH24ateuXerQoYPHsXbt2mnXrl3VDgUAZhz+7jutGT1arjNnZBiGfklN1VWvvaZG5fxbBSA4mJ7JCQsL06FDhzyOlZwEDAA1xTAMrZs4Uc6CgrNNjtMp56lT2vDHP/o7GgA/M93k9OnTR08//XSZlY3z8/P117/+1X2HcACoCWdOnlRhTo507hpZLpfyMzLkLCz0XzAAfmd62uXJJ59Ut27d1KpVKw0ZMkRNmzZVRkaG3nnnHZ0+fVpLly71RU4A8MgRGSl7aKhcZ86U2h4SFqaQcs4fBBAcTM/kXHLJJdqyZYuuuuoqvfvuu5o+fbreffddXX311dq8ebMuueQSX+QEAI/sISG6dNIk6dzFw2w2dZ48OagWUANQlqmZHMMwdOTIETVr1kyLFy/2VSYAMKXtyJEKj43VnlWrVHT6tC6+5RZddN497AAEH1NNjtPpVHx8vN5//31df/31vsoEAKZddP31SrruOmVlZXm8Lx2A4GPqcFVoaKiaNWum4uJiX+UBAADwCtPn5Nx///164YUXVFRU5Is8AAAAXmH66qq9e/fqp59+UlJSkvr27au4uLhSJ/fZbDbuKQUAAPzOdJOzcuVK920dNm/eXGacJgcAAAQC001Oenq6L3IAAAB4lelzcgAAAGqDKjU5hw8f1qOPPqqrrrpKbdq00Q8//CBJmjt3rr766iuvBgQAAKgK003Ot99+q9atW2vJkiWKj4/X7t273VdaHTx4UC+88ILXQwIAAJhlusn54x//qG7dumn37t16/fXXZZxzU7zf/e53zOQAAICAYPrE4y1btmjFihUKDQ0tsyhgbGyssrOzvRYOAACgqkzP5ERGRurEiRMex/bt26dGjRpVOxQAAEB1mW5yrrnmGs2cOVO5ubnubTabTQUFBZo7d64GDBjg1YAAcCGGYSj9ww/1+T336PvHH9fejz/2dyQAAcD04apnnnlGPXr0UOvWrXXllVfKZrNpypQp+vHHH2Wz2TRz5kxf5ASAcu1YuFDfzZkjw+WSJG385hsVHT2qS26/3c/JAPiT6ZmcZs2a6b///a8eeOABZWZmqlWrVsrNzdXw4cP19ddfKy4uzhc5AcAjV3GxvnvxRXeDc3aj62zTc86FEQCCj+mZHEmKjo5WSkqKUlJSvJ0HAExx5ufLcDrLbC8uLFRxUZEcYWF+SAUgEFS6yfnhhx+0YMECpaenq2nTprr11lvVv39/X2YDgAsKrVdP4XFxKsjJkf43c2Oz2xXRtCkNDhDkKnW4asOGDerSpYtSU1O1efNmLVy4UNdcc43mz5/v63wAUCGbzaZec+cqNDJS9tBQ2RwOOaKi1GvOHH9HA+BnlZrJmT59utq1a6cPP/xQiYmJOnHihEaNGqUpU6bonnvu8XVGAKhQ406dNHDlSmVt3qzjx4+rzdVXKyI21t+xAPhZpZqc77//XvPnz1diYqIkqX79+nruuefUsmVL7d+/370dAPwlPDZWydddp6ysLIWxXhcAVfJw1eHDh9W8efNS20oam8OHD3s/FQAAQDVV+hJym83myxwAAABeVemrq6688krZ7WV7ol69epXabrPZdPz4ce+kAwAAqKJKNTnTpk3zdQ4AAACvoskBAACWZPq2DgAAALVBlW7rAACBxDAM7fnoI6WvXKnTp0+raMgQtRg40N+xAPgZTQ6AWo+7kAPwJCibnKioKDkcDo93KDYMwz0WDHcwLvmMwfBZJeprRRXdhbzN8OGWXv4iGOp7LvZfa7tQfR0O8y1LUDY5nTt3VkxMjJwe7lwsSTExMXK5XHKd+4+mxRUXF/s7Qo2hvtZy+sSJcu9Cfjo/XyFBcJNOK9f3fOy/1lZRfWNiYky/XlA2OVu3blXHjh0V6+HeNi6XS7m5uWrUqJHHdYGsxjAMFRcXKyQkxNK/8ZagvtYTEhNT7l3I60ZF+TmdbwVDfc/F/mttF6pvTk6O6dcMyiYnLy9PTqfT418am83mHguGv1QlguXzUl/rKbkL+doxY1RcVHR2yjsiQr3mzLHsZz6flet7LvZfa7tQfcs7+lKRoGxyAFgLdyEH4AlNDgBL4C7kAM5n/YOaAAAgKNHkAAAAS6LJAQAAlkSTAwAALIkmBwAAWBJNDgAAsCSaHAAAYEk0OQAAwJJocgAAgCXR5AAAAEuiyQEAAJZEkwMAACyJJgcAAFgSTQ4AALAkmhwAAGBJNDkAAMCSaHIAAIAl0eQAAABLoskBAACWRJMDAAAsiSYHAABYEk0OAACwJJocAABgSTQ5AADAkmhyAACAJdHkAAAAS6LJAQAAlkSTAwAALIkmBwAAWBJNDgAAsCSaHAAAYEk0OQAAwJIc/g5QXXl5eUpNTdW3336r8PBwDR06VAMGDPB3LAAA4Ge1vslZsGCBiouLtWjRImVmZuovf/mLmjdvrk6dOvk7GgAA8KNa3eQUFhbqyy+/1Jw5cxQREaFWrVqpX79+Wr16NU0OEETOnDmjj66+WoU5OZKk8Ph4DUpL83MqAP5Wq8/JOXjwoCQpKSnJva1ly5bau3evvyIB8IMV3bq5GxxJKsjK0ttXXOHHRAACQa2fyQkPDy+1LTIyUgUFBaW2ZWZmKjMzU5KUk5Oj/Px8SZLL5SrzmiXbPI1ZkWEYcrlccrlcstls/o7jc9TXeoqOH1fxefu8JDnz81VUVKTQ0FA/pKoZwVDfc7H/Wpsv6lurm5ywsLAyDU1+fn6ZxmfBggVKSUlxPx42bJgkKSsrq9zXzs7O9mJSBBrqax2Ht2wpd2zP11+rXosWNZgGNYH919q8Wd9a3eQ0a9ZMkrR//34lJiZKktLT05WcnFzqeePGjdONN94o6exMzurVqyVJ8fHxZV7T5XIpOztbcXFxsttr9dG8SjEMQ06nUw6HI2h+U6C+1tLo2mu1/fHHPY617tathtPUrGCo77nYf63tQvWtaGKiPLW6yQkLC1OPHj20ePFiTZgwQYcOHVJaWpoefvjhUs9LSEhQQkKCJCkjI0MbN26UpAp3ErvdHjQ7UclnDYadqAT1tY66deuqaZ8+yvjii1LbE6+5xvI1Dob6esL+a23erG+tbnKks7M0L730kkaOHKmIiAgNHz5cl156qb9jAahBfV9+WT+//ba2L1ggl8ulSydM0G8GDfJ3LAB+VuubnKioKD366KP+jgHAz1oPHapWQ4YoKyvL46FoAMHH+vN9AAAgKNHkAAAAS6LJAQAAlkSTAwAALIkmBwAAWBJNDgAAsCSaHAAAYEk0OQAAwJJocgAAgCXR5AAAAEuiyQEAAJZU6+9dVVWHDx+ucLwqt3SvjRwOh2JiYpSTkyOn0+nvODWG+lob9bU26mtt5dX3Qj+3PQm6JiciIkKhoaFasWKFx/GTJ0/qm2++UdeuXVWvXr0aTgdfo77WRn2tjfpaW2XqGxoaqoiIiEq/ps0wDMNbAWuLY8eO6dSpUx7Htm3bpmuvvVaffvqpOnbsWMPJ4GvU19qor7VRX2urTH0jIiIUHR1d6dcMupkcSYqOji73f1LJNFlsbKyaNm1ak7FQA6ivtVFfa6O+1uaL+nLiMQAAsKSQ6dOnT/d3iEATFRWlvn37cszXoqivtVFfa6O+1ubt+gblOTkAAMD6OFwFAAAsiSYHAABYUlBeXXW+f/7zn/rkk0/kcrnUq1cvjR07Vg5Hxf9rtm3bpj//+c+65ZZbdOedd9ZQUlRFZet78OBB/eMf/9DOnTvldDqVnJysu+++W61bt/ZDapiVl5en1NRUffvttwoPD9fQoUM1YMAAf8eCSZWt45YtW/Tuu+9q7969cjgc6tSpk0aPHq2YmBg/pIYZVdlXlyxZoqVLl2ratGnq2rVrpd8r6Gdy/vWvf2ndunV6/vnnNX/+fP366696++23K/yeM2fO6NVXX9Ull1xSQylRVWbqm5+fr65duyo1NVX//Oc/1bNnT6WkpKiwsLCGU6MqFixYoOLiYi1atEhTp07V4sWL9f333/s7FkyqbB1PnTqlW265RYsWLdIrr7wiu92uuXPn+iExzDK7rx48eFAbN25Uw4YNTb9X0Dc5q1ev1s0336wmTZqoQYMGGjp0qFavXl3h9yxfvlxXXHEF6zTUAmbq26ZNG1177bWqX7++QkJCNHDgQBUWFurAgQM1nBpmFRYW6ssvv9SIESMUERGhVq1aqV+/fhfclxFYzNSxT58++u1vf6uwsDCFh4frhhtu0I4dO/yQGmZUZV99+eWXddddd13wCIsnQd/k7Nu3TxdddJH7cYsWLXT48GHl5+d7fP7Bgwe1fv163XrrrTWUENVhtr7n+vnnn2UYhhISEnyYEN5w8OBBSVJSUpJ7W8uWLbV3715/RUIVVKeO27dvL/V9CExma5yWlqYGDRqoc+fOVXo/Szc5xcXFFf6RznaVkZGR7u8p+bqgoMDja7788ssaNWqU6tat6/sPgAr5or4lTpw4oeeff17Dhw8v9f0ITIWFhQoPDy+1LTIy8oJ1RmCpah137dql5cuXa9SoUb6MBy8wU+MTJ05o6dKluvvuu6v8fpY+8Xjq1Knavn27x7Ho6Gi98cYbCgsLK3Ufq5Kvzy+CJK1Zs0bh4eG64oorfBMYpni7viXy8/OVkpKiLl26aPDgwd4NDZ8ICwsr849kfn5+hXVG4KlKHdPT0zVz5kzdf//9ateuna8joprM1HjRokUaOHCgGjVqVOX3s3ST8+STT17wOUlJSUpPT1fbtm0lnd1hGjdu7PG39++//17btm3T8OHDJZ3tSG02m3bt2qVZs2Z5NzwuyNv1lc42QdOmTVOrVq00ZswYr+aF7zRr1kyStH//fiUmJko6W+vk5GR/xoJJZuu4d+9eTZs2TXfddZd69OhRYzlRdWZqvG3bNm3atMl9sUh+fr5mz56tgQMHasSIEZV6P0s3OZVx1VVX6b333tPll1+usLAwLVu2TP379/f43NGjR5f6H/vaa6+pQYMGGjZsWE3FhUlm6lvS4CQmJuree++t4aSojrCwMPXo0UOLFy/WhAkTdOjQIaWlpenhhx/2dzSYYKaO+/bt09SpU3XHHXeob9++NR8WVWKmxrNnz3afeiBJDz/8sG6//XZTR1OC/rYOhmFo8eLF+uSTT1RcXKzevXuXWkflvvvu06233upxJ5ozZ45iYmJYJyeAmalvWlqa5s6dq7p168pms7lfY9q0aWrfvr2/PgIqKS8vTy+99JK+/fZbRUREsE5OLVVeHXNycnTfffcpNTVVsbGxmjt3rtasWVPm/MgLLQEC/6tsjc83evRo3XvvvabWyQn6JgcAAFiTpa+uAgAAwYsmBwAAWBJNDgAAsCSaHAAAYEk0OQAAwJJocgAAgCXR5AAAAEuiyQEAAJZEkwMAACyJJgcAAFgSTQ4AALAkmhwAAGBJ/z/e1h1XMlofAAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (283332630)>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pln.ggplot(data=df_win_per, mapping=pln.aes(x='0',y='Wper'))+pln.geom_boxplot(colour=\"brown\", fill=\"orange\")+pln.theme_bw()+pln.xlab(\"\")+pln.ylab(\"Percentage of winner's victories\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So in average, the percentage of victories for a winner in the past is equal in average to about **62%** which makes sense since its > 50%.  \n",
    "Still, even if the results show that more than 75% (3rd quartile) of the cases have a winning percentage > 50%, some observations lay bellow this tresh-hold (about 2%). One explanation to this is the absence of complete scores for some matches, or the player withdrawal/retirement during the match, the 'Comment' feature can be used to check this hypothesis.     \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Abberant observations : 0.022\n"
     ]
    }
   ],
   "source": [
    "print(\"Abberant observations : \"+ str(round(sum(df_win_per.Wper<50)/len(df_win_per), 3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-------\n",
    "**Aditional Part**  \n",
    "If we call 'victory' during a given match a **Set** won in that match, then the percentage of victories of the winner in the past is expressed as the ratio :  \n",
    "\\begin{equation*}\n",
    "\\frac{Wsets}{Wsets+Lsets}\n",
    "\\end{equation*}  \n",
    "In this case, we'll follow almost the same steps as before"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Select the columns names\n",
    "columns_2 = ['Wsets','Lsets']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_sub_2 = df_atp.loc[:,columns_2].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wsets</th>\n",
       "      <th>Lsets</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Wsets Lsets\n",
       "0    2.0     0\n",
       "1    2.0     0\n",
       "2    2.0     1\n",
       "3    2.0     0\n",
       "4    2.0     1"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_sub_2.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Find indexes with missing values in the first set (W1 or L1)\n",
    "naIndexes_2 = df_atp_sub_2['Wsets'].isna() | df_atp_sub_2['Lsets'].isna()\n",
    "cleanIndexes_2 = naIndexes_2.apply(lambda e : not e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "51951"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(cleanIndexes_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "52174"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(df_atp_sub_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Filter data with cleanIndexes\n",
    "df_atp_sub_2 = df_atp_sub_2.loc[cleanIndexes_2,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 1, 2, '`1', 3.0], dtype=object)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Unique values in Li features (i in 1:5)\n",
    "pd.concat([df_atp_sub_2.Lsets, df_atp_sub_2.Wsets]).unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function : handles missing values/special caracters + calculates the winning percentage\n",
    "def WinPercentage2(obs):\n",
    "    myobs = obs.copy()\n",
    "    myobs[myobs == '`1'] = 1\n",
    "    wins = np.nansum(myobs['Wsets'])\n",
    "    loses = np.nansum(myobs['Lsets'])\n",
    "    if wins+loses == 0:\n",
    "        return(0)\n",
    "    else:\n",
    "        per = 100*wins/(wins+loses)\n",
    "        return(per)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Execution will take about 2 minutes\n",
    "df_win_per_2 = df_atp_sub_2.apply(axis=1, func=WinPercentage2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wper</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>51951.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>86.939199</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>17.906833</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>66.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               Wper\n",
       "count  51951.000000\n",
       "mean      86.939199\n",
       "std       17.906833\n",
       "min        0.000000\n",
       "25%       66.666667\n",
       "50%      100.000000\n",
       "75%      100.000000\n",
       "max      100.000000"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_win_per_2 = pd.DataFrame({'Wper':df_win_per_2})\n",
    "df_win_per_2.describe(include='all')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (283450197)>"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pln.ggplot(data=df_win_per_2, mapping=pln.aes(x='Wper'))+pln.geom_density(colour=\"brown\", fill=\"orange\")+pln.theme_bw()+pln.xlab(\"\")+pln.ylab(\"Density\")+pln.xlab(\"Percentage of winner's victories\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (283457968)>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pln.ggplot(data=df_win_per, mapping=pln.aes(x='0',y='Wper'))+pln.geom_boxplot(colour=\"brown\", fill=\"orange\")+pln.theme_bw()+pln.xlab(\"\")+pln.ylab(\"Percentage of winner's victories\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When comparing Sets results to single games, we get more insights, let's look at both distribution informations :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wper</th>\n",
       "      <th>Wper</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>51914.000000</td>\n",
       "      <td>51951.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>61.913378</td>\n",
       "      <td>86.939199</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>8.635817</td>\n",
       "      <td>17.906833</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>56.000000</td>\n",
       "      <td>66.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>60.000000</td>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>66.666667</td>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>100.000000</td>\n",
       "      <td>100.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               Wper          Wper\n",
       "count  51914.000000  51951.000000\n",
       "mean      61.913378     86.939199\n",
       "std        8.635817     17.906833\n",
       "min        0.000000      0.000000\n",
       "25%       56.000000     66.666667\n",
       "50%       60.000000    100.000000\n",
       "75%       66.666667    100.000000\n",
       "max      100.000000    100.000000"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([df_win_per,df_win_per_2], axis=1).describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Match winners win on average **87% of the Sets** but just **62% of the Games**. A simple explanation for that is the scoring system : you have to win 6 games to get 1 Set, we can look at the Set result as an average of the 6 (or more) played games, and thus it represents the overall performance of the player, while single games are 'less' general and thus they fluctuate more, and generally a player can't win many consecutive games. Besides, another explanation is that Set results are somehow more crutial since they are directly linked to match final result, while games results come in a second level far from the final result.  \n",
    "**hint**  \n",
    "This issue of ensemble probability Vs time probability reminds me of an article by Nassim Nicholas Taleb on Medium. You can check it if you want to : https://medium.com/incerto/the-logic-of-risk-taking-107bf41029d3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "****************************"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Preparation\n",
    "## Data Splitting"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We start first by selecting the rows we're going to use, they belong to 4 categories :  \n",
    "1 - Location and organisation related features ( Tournament, Series, Date, Round)  \n",
    "2 - Game field related data (Surface, Court)  \n",
    "3 - Players' performances (LPts, WPts, WRank, LRank)  \n",
    "4 - Average betting data (AvgW, AvgL)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "usedColumns = ['Best of',\n",
    "'Court',\n",
    "'Date',\n",
    "'LPts',\n",
    " 'LRank',\n",
    "'Loser',\n",
    " 'Round',\n",
    " 'Series',\n",
    " 'Surface',\n",
    " 'Tournament',\n",
    "'WPts',\n",
    " 'WRank',\n",
    " 'Winner']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Preparing rows indexes for splitting (Train-Test)\n",
    "gTest = lambda x: x.year == 2017\n",
    "gTrain = lambda x: x.year < 2017\n",
    "testInd = df_atp['Date'].apply(gTest)\n",
    "trainInd = df_atp['Date'].apply(gTrain)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of atp data : a)Train = (47028, 13) ; b)Test = (2635, 13)\n"
     ]
    }
   ],
   "source": [
    "#Splitting data (Train - Test)\n",
    "df_atp_Train = df_atp.loc[trainInd, usedColumns].copy()\n",
    "df_atp_Test = df_atp.loc[testInd, usedColumns].copy()\n",
    "print(\"Shape of atp data : a)Train = \"+str(df_atp_Train.shape)+ \" ; b)Test = \"+str(df_atp_Test.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>47028.00000</td>\n",
       "      <td>47028</td>\n",
       "      <td>47028</td>\n",
       "      <td>30772.000000</td>\n",
       "      <td>46957.0</td>\n",
       "      <td>47028</td>\n",
       "      <td>47028</td>\n",
       "      <td>47028</td>\n",
       "      <td>47028</td>\n",
       "      <td>47028</td>\n",
       "      <td>30824.000000</td>\n",
       "      <td>47013.0</td>\n",
       "      <td>47028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unique</th>\n",
       "      <td>NaN</td>\n",
       "      <td>2</td>\n",
       "      <td>4087</td>\n",
       "      <td>NaN</td>\n",
       "      <td>924.0</td>\n",
       "      <td>1385</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "      <td>4</td>\n",
       "      <td>206</td>\n",
       "      <td>NaN</td>\n",
       "      <td>586.0</td>\n",
       "      <td>849</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>top</th>\n",
       "      <td>NaN</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-05-29 00:00:00</td>\n",
       "      <td>NaN</td>\n",
       "      <td>46.0</td>\n",
       "      <td>Lopez F.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>French Open</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Federer R.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>freq</th>\n",
       "      <td>NaN</td>\n",
       "      <td>38521</td>\n",
       "      <td>127</td>\n",
       "      <td>NaN</td>\n",
       "      <td>415.0</td>\n",
       "      <td>370</td>\n",
       "      <td>21848</td>\n",
       "      <td>12204</td>\n",
       "      <td>24710</td>\n",
       "      <td>2159</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1077.0</td>\n",
       "      <td>1023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>first</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>2000-01-03 00:00:00</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>last</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>2016-11-20 00:00:00</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>3.37097</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1033.792604</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1845.345672</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.77739</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1222.460585</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>2360.801597</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>3.00000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>3.00000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>490.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>631.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3.00000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>726.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1000.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.00000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1125.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1875.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>5.00000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>16950.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>16950.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Best of    Court                 Date          LPts    LRank  \\\n",
       "count   47028.00000    47028                47028  30772.000000  46957.0   \n",
       "unique          NaN        2                 4087           NaN    924.0   \n",
       "top             NaN  Outdoor  2000-05-29 00:00:00           NaN     46.0   \n",
       "freq            NaN    38521                  127           NaN    415.0   \n",
       "first           NaN      NaN  2000-01-03 00:00:00           NaN      NaN   \n",
       "last            NaN      NaN  2016-11-20 00:00:00           NaN      NaN   \n",
       "mean        3.37097      NaN                  NaN   1033.792604      NaN   \n",
       "std         0.77739      NaN                  NaN   1222.460585      NaN   \n",
       "min         3.00000      NaN                  NaN      1.000000      NaN   \n",
       "25%         3.00000      NaN                  NaN    490.000000      NaN   \n",
       "50%         3.00000      NaN                  NaN    726.000000      NaN   \n",
       "75%         3.00000      NaN                  NaN   1125.000000      NaN   \n",
       "max         5.00000      NaN                  NaN  16950.000000      NaN   \n",
       "\n",
       "           Loser      Round         Series Surface   Tournament          WPts  \\\n",
       "count      47028      47028          47028   47028        47028  30824.000000   \n",
       "unique      1385          8              8       4          206           NaN   \n",
       "top     Lopez F.  1st Round  International    Hard  French Open           NaN   \n",
       "freq         370      21848          12204   24710         2159           NaN   \n",
       "first        NaN        NaN            NaN     NaN          NaN           NaN   \n",
       "last         NaN        NaN            NaN     NaN          NaN           NaN   \n",
       "mean         NaN        NaN            NaN     NaN          NaN   1845.345672   \n",
       "std          NaN        NaN            NaN     NaN          NaN   2360.801597   \n",
       "min          NaN        NaN            NaN     NaN          NaN      1.000000   \n",
       "25%          NaN        NaN            NaN     NaN          NaN    631.000000   \n",
       "50%          NaN        NaN            NaN     NaN          NaN   1000.000000   \n",
       "75%          NaN        NaN            NaN     NaN          NaN   1875.000000   \n",
       "max          NaN        NaN            NaN     NaN          NaN  16950.000000   \n",
       "\n",
       "          WRank      Winner  \n",
       "count   47013.0       47028  \n",
       "unique    586.0         849  \n",
       "top         1.0  Federer R.  \n",
       "freq     1077.0        1023  \n",
       "first       NaN         NaN  \n",
       "last        NaN         NaN  \n",
       "mean        NaN         NaN  \n",
       "std         NaN         NaN  \n",
       "min         NaN         NaN  \n",
       "25%         NaN         NaN  \n",
       "50%         NaN         NaN  \n",
       "75%         NaN         NaN  \n",
       "max         NaN         NaN  "
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.describe(include='all')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#  Data Preprocessing "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.preprocessing import PolynomialFeatures\n",
    "\n",
    "#Import Random Forest Model\n",
    "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor\n",
    "from sklearn.model_selection import RandomizedSearchCV"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1 - Missing Values Imputation\n",
    "### 1.1 - Overview of data missing values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We start by examining columns with missing values. The columns with the most NA's are those Points related."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percentage of missing values in WPts column : 34.456%\n",
      "Percentage of missing values in LPts column : 34.567%\n"
     ]
    }
   ],
   "source": [
    "print(\"Percentage of missing values in WPts column : \" + str(np.round(100*np.sum(df_atp_Train.WPts.isna())/len(df_atp_Train), 3))+\"%\")\n",
    "print(\"Percentage of missing values in LPts column : \" + str(np.round(100*np.sum(df_atp_Train.LPts.isna())/len(df_atp_Train), 3))+\"%\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A quick exploration of the unique values in 'WRank' and 'LRank' shows that there are 3 types of values : **Numerical, 'NR', and nan**. We first turn 'NR' (*Not recognized*) to np.NaN to deal with it as a missing value."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "naRankIndW = df_atp_Train.loc[:,'WRank'].apply(lambda x: x=='NR') \n",
    "naRankIndL = df_atp_Train.loc[:,'LRank'].apply(lambda x: x=='NR')\n",
    "df_atp_Train.loc[naRankIndW,'WRank'] = np.NaN\n",
    "df_atp_Train.loc[naRankIndL,'LRank'] = np.NaN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in WRank column : 16\n",
      "Number of missing values in LRank column : 98\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in WRank column : \" + str(np.sum(df_atp_Train.WRank.isna())))\n",
    "print(\"Number of missing values in LRank column : \" + str(np.sum(df_atp_Train.LRank.isna())))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Fortunately, 'Date' column is clear with no missing values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in Date column : 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in Date column : \" + str(np.sum(df_atp_Train.Date.isna())))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 - Imputation of Rank-related missing values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We start by imputing the Rank-related missing values since they are less frequent, using a per-player linear regression model modeling Rank = f(Date). We'll use these values later to impute Points-related values since there is a large correlation between players' rankings and points (Ranking is by definition based on Points).  \n",
    "To do so, we distinguish two separate cases :  \n",
    "- **Players having some non-NA (date,rank) observations** : in this case, we will fit a simple linear regression model to interpolate the missing ranks.  \n",
    "- **Players having no recorded (date,rank) observation** : in this case, we will impute the median value of winners' ranks and losers' rank for winner and loser cases respectively.  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll try to see some per-player Date - Rank relationship."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "myPlayer = 'Federer R.'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "winData = df_atp_Train.loc[df_atp_Train.Winner==myPlayer,['Date', 'WRank']]\n",
    "loseData = df_atp_Train.loc[df_atp_Train.Loser==myPlayer,['Date','LRank']]\n",
    "    \n",
    "RankData = pd.DataFrame({'Date':pd.concat([winData.Date, loseData.Date], axis=0), \n",
    "                             'Rank':pd.concat([winData.WRank, loseData.LRank], axis=0)})\n",
    "RankData.sort_index(axis=0, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "RankData['Date'] = RankData['Date'].apply(lambda x : x.year +(x.month-1)/12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "RankData.dropna(inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (281356243)>"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pln.ggplot(RankData, pln.aes(x='Date', y='Rank'))+pln.geom_point(pln.aes(colour='Rank'))+pln.theme_bw()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The shape of the time serie varies of course from a player to another one. One simple way to model this relationship is using a polynomial regression with a degree depending on the size of available data of the given player."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights_date_rank = np.polyfit(RankData.Date, RankData.Rank, 3)\n",
    "model_date_rank = np.poly1d(weights_date_rank)\n",
    "preds_rank = np.round(model_date_rank(RankData.Date))\n",
    "preds_rank[preds_rank<1] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (-9223372036572904967)>"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ggDateRankex = pln.ggplot(data = pd.DataFrame({'Date':RankData.Date,'Predicted Rank':preds_rank}))\n",
    "ggDateRankex + pln.geom_point(pln.aes(x='Date',y='Predicted Rank',colour= 'Predicted Rank'))+ pln.theme_bw()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The fitted model seems to be close enough to the true shape. We will use it for the whole set of players to impute the missing values of ranks."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Players having missing ranks:\n",
    "df_nan_RanksW = df_atp_Train.loc[df_atp_Train.WRank.isna(),['Date','Winner']].copy() \n",
    "df_nan_RanksL = df_atp_Train.loc[df_atp_Train.LRank.isna(),['Date','Loser']].copy() \n",
    "Trues0 = np.repeat(True, repeats=df_nan_RanksW.shape[0], axis=0)\n",
    "Falses0 = np.repeat(False, repeats=df_nan_RanksL.shape[0], axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Dataframe putting together all the players with NA's in Rank + Date of match + whether it's a winner or not\n",
    "df_nan_Ranks = pd.DataFrame({'Date':pd.concat([df_nan_RanksW['Date'], df_nan_RanksL['Date']], axis=0),\n",
    "                          'Player' : pd.concat([df_nan_RanksW['Winner'],df_nan_RanksL['Loser']], axis=0),\n",
    "                         'isWinner':pd.Series(np.concatenate([Trues0,Falses0]).flatten(), \n",
    "                                            index=np.concatenate([df_nan_RanksW.index, df_nan_RanksL.index]))\n",
    "                         })\n",
    "df_nan_Ranks.sort_index(ascending=True, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_nan_Ranks['DateFloat'] = df_nan_Ranks['Date'].apply(lambda x : x.year + (x.month-1)/12 + x.day/365)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Date</th>\n",
       "      <th>Player</th>\n",
       "      <th>isWinner</th>\n",
       "      <th>DateFloat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>63</th>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>Al-Alawi S.K.</td>\n",
       "      <td>False</td>\n",
       "      <td>2000.008219</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>377</th>\n",
       "      <td>2000-02-14</td>\n",
       "      <td>Jensen L.</td>\n",
       "      <td>False</td>\n",
       "      <td>2000.121689</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>560</th>\n",
       "      <td>2000-02-28</td>\n",
       "      <td>Roddick A.</td>\n",
       "      <td>False</td>\n",
       "      <td>2000.160046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>611</th>\n",
       "      <td>2000-03-06</td>\n",
       "      <td>Gonzalez P.</td>\n",
       "      <td>False</td>\n",
       "      <td>2000.183105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>613</th>\n",
       "      <td>2000-03-06</td>\n",
       "      <td>Hadad M.</td>\n",
       "      <td>False</td>\n",
       "      <td>2000.183105</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Date         Player  isWinner    DateFloat\n",
       "63  2000-01-03  Al-Alawi S.K.     False  2000.008219\n",
       "377 2000-02-14      Jensen L.     False  2000.121689\n",
       "560 2000-02-28     Roddick A.     False  2000.160046\n",
       "611 2000-03-06    Gonzalez P.     False  2000.183105\n",
       "613 2000-03-06       Hadad M.     False  2000.183105"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_nan_Ranks.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Winners' median rank : 41.0\n",
      "Losers' median rank : 65.0\n"
     ]
    }
   ],
   "source": [
    "#Preparing median values for rank-imputations in case 2 : \n",
    "medianRankW = np.median(df_atp_Train['WRank'])\n",
    "print(\"Winners' median rank : \"+str(medianRankW))\n",
    "medianRankL = np.median(df_atp_Train['LRank'])\n",
    "print(\"Losers' median rank : \"+str(medianRankL))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "def genMedianRank(lenWin, lenLose, medianRankW = medianRankW, medianRankL =medianRankL):\n",
    "    \"\"\"\n",
    "    -Overview : \n",
    "    Rank Imputation function for case 2 players\n",
    "    \n",
    "    - Arguments : \n",
    "    lenWin , lenLose : number of won/lost matches for a given player \n",
    "    medianRankW , medianRankL :  Rank medians (Winner/Loser)\n",
    "    \n",
    "    - Returns :\n",
    "    numpy array of juxtaposed Rank medians (Winner/Loser) . This will serve as an imputation vector (shape : (lenWin+lenLose))\n",
    "    \"\"\"\n",
    "    return(np.concatenate([np.repeat(medianRankW, repeats=lenWin, axis=0),\n",
    "                           np.repeat(medianRankL, repeats=lenLose, axis=0)]).flatten())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In order to handle Rank missing values in the test set, we're going to use the same per player fitted model "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeRankImput(Player, nanData = df_nan_Ranks ,sourceData = df_atp_Train):\n",
    "    \"\"\"\n",
    "    - Overview :\n",
    "    Rank Imputations Computing function for both cases (1 & 2)\n",
    "    \n",
    "    - Arguments :\n",
    "    Player : [str] Player's name\n",
    "    nanData : [pandas DataFrame] data of players with Rank missing values (plays the role of training data set)\n",
    "    sourceData : [pandas DataFrame] target dataset to compute imputations (plays the role of test data set)\n",
    "    \n",
    "    - Returns : \n",
    "    [pandas Data Frame] : subset of nanData for the given Player + computed imputation values\n",
    "    \"\"\"\n",
    "    \n",
    "    winData = sourceData.loc[sourceData.Winner==Player,['Date', 'WRank']]\n",
    "    loseData = sourceData.loc[sourceData.Loser==Player,['Date','LRank']]\n",
    "    \n",
    "    playerNanData = nanData.loc[nanData['Player']==Player,:].copy()\n",
    "    \n",
    "    RankData = pd.DataFrame({'Date':pd.concat([winData.Date, loseData.Date], axis=0), \n",
    "                             'Rank':pd.concat([winData.WRank, loseData.LRank], axis=0)})\n",
    "    RankData.sort_index(inplace=True)\n",
    "    RankData.dropna(inplace=True)\n",
    "    RankData['Date'] = RankData['Date'].apply(lambda x : x.year + (x.month-1)/12)\n",
    "    \n",
    "    if(RankData.shape[0]==0):\n",
    "        return(pd.concat([playerNanData, pd.DataFrame({'PredRank':genMedianRank(lenWin=len(winData),lenLose=len(loseData))},\n",
    "                                                      index=playerNanData.index)], axis=1))\n",
    "    \n",
    "    indToDrop = RankData['Rank'].apply(lambda x : str(x).isalpha())\n",
    "    indToDrop = indToDrop.index[indToDrop]\n",
    "    RankData.drop(axis=0,index= indToDrop, inplace=True)\n",
    "    \n",
    "    polyDegree = round(np.log10(len(RankData)))     #polynomial degree depending on how much data we have (log like for example)\n",
    "        \n",
    "    weights_date_rank = np.polyfit(RankData.Date, RankData.Rank, polyDegree)\n",
    "    model_date_rank = np.poly1d(weights_date_rank)\n",
    "    \n",
    "    prediction = np.round(model_date_rank(playerNanData['DateFloat']))\n",
    "    prediction[prediction<1] = 1\n",
    "    \n",
    "    result = pd.concat([playerNanData, pd.DataFrame({'PredRank':prediction}, index=playerNanData.index)], axis=1)\n",
    "    return(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "def imputeRankData(nanData = df_nan_Ranks ,sourceData = df_atp_Train, targetData = df_atp_Train):\n",
    "    \"\"\"\n",
    "    - Overview :\n",
    "    Inplace Rank Imputation of data passed as sourceData based on computed imputations by the previous function 'computeRankImput'\n",
    "    \n",
    "    -Arguments :\n",
    "    nanData : [pandas DataFrame] data of players with Rank missing values\n",
    "    sourceData : [pandas DataFrame] target dataset to be imputed\n",
    "    \n",
    "    -Returns:\n",
    "    Inplace imputations (no return)\n",
    "    Prints imputations progress\n",
    "    \"\"\"\n",
    "    progress = 0\n",
    "    nanPtsLen = len(nanData['Player'].unique())\n",
    "    tenPer, per = nanPtsLen//10, 0\n",
    "    for p in nanData['Player'].unique():\n",
    "        imputations = computeRankImput(Player=p, nanData = nanData ,sourceData = sourceData)        \n",
    "        for i,val in enumerate(imputations['isWinner']):\n",
    "            imp = imputations.loc[imputations.index[i], 'PredRank']\n",
    "            if (val): #Winner case\n",
    "                targetData.loc[imputations.index[i],'WRank'] = imp\n",
    "            else:     #Loser case\n",
    "                targetData.loc[imputations.index[i],'LRank'] = imp\n",
    "        if(per==progress):\n",
    "            print(\"Imputations in progress \"+str( np.round(100*per/nanPtsLen, decimals=0))+\"%\")\n",
    "            per += tenPer\n",
    "        progress += 1\n",
    "        if(progress==nanPtsLen):\n",
    "            print(\"Done !\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Imputations in progress 0.0%\n",
      "Imputations in progress 9.0%\n",
      "Imputations in progress 19.0%\n",
      "Imputations in progress 28.0%\n",
      "Imputations in progress 37.0%\n",
      "Imputations in progress 47.0%\n",
      "Imputations in progress 56.0%\n",
      "Imputations in progress 65.0%\n",
      "Imputations in progress 74.0%\n",
      "Imputations in progress 84.0%\n",
      "Imputations in progress 93.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "imputeRankData(nanData = df_nan_Ranks ,sourceData = df_atp_Train, targetData = df_atp_Train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in WRank column : 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in WRank column : \" + str(sum(df_atp_Train.WRank.isna())))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in LRank column : 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in LRank column : \" + str(sum(df_atp_Train.LRank.isna())))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train['WRank'] = df_atp_Train['WRank'].astype('int64')\n",
    "df_atp_Train['LRank'] = df_atp_Train['LRank'].astype('int64')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We do the same for the test data set. This time, we can use even Ranks from Test Data to perform the interpolation (we couldn't do that before because we risk a Data Leakage).  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "naRankIndWTest = df_atp_Test.loc[:,'WRank'].apply(lambda x: x=='NR') \n",
    "naRankIndLTest = df_atp_Test.loc[:,'LRank'].apply(lambda x: x=='NR')\n",
    "df_atp_Test.loc[naRankIndWTest,'WRank'] = np.NaN\n",
    "df_atp_Test.loc[naRankIndLTest,'LRank'] = np.NaN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in WRank column (Test data): 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in WRank column (Test data): \" + str(sum(df_atp_Test.WRank.isna())))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in LRank column (Test data): 6\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in LRank column (Test data): \" + str(sum(df_atp_Test.LRank.isna())))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Players having missing ranks (In test):\n",
    "df_nan_RanksL_T = df_atp_Test.loc[df_atp_Test.LRank.isna(),['Date','Loser']].copy() \n",
    "Falses0T = np.repeat(False, repeats=df_nan_RanksL_T.shape[0], axis=0)\n",
    "\n",
    "#Dataframe putting together all the players with NA's in Rank + Date of match + whether it's a winner or not\n",
    "df_nan_Ranks_T = pd.DataFrame({'Date':df_nan_RanksL_T['Date'],\n",
    "                          'Player' : df_nan_RanksL_T['Loser'],\n",
    "                         'isWinner':pd.Series(Falses0T, index=df_nan_RanksL_T.index)\n",
    "                         })\n",
    "df_nan_Ranks_T.sort_index(ascending=True, inplace=True)\n",
    "\n",
    "df_nan_Ranks_T['DateFloat'] = df_nan_Ranks_T['Date'].apply(lambda x : x.year + (x.month-1)/12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Date</th>\n",
       "      <th>Player</th>\n",
       "      <th>isWinner</th>\n",
       "      <th>DateFloat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47124</th>\n",
       "      <td>2017-01-10</td>\n",
       "      <td>Sitak A.</td>\n",
       "      <td>False</td>\n",
       "      <td>2017.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47225</th>\n",
       "      <td>2017-01-17</td>\n",
       "      <td>Haas T.</td>\n",
       "      <td>False</td>\n",
       "      <td>2017.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47592</th>\n",
       "      <td>2017-02-28</td>\n",
       "      <td>Alawadhi O.</td>\n",
       "      <td>False</td>\n",
       "      <td>2017.083333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48235</th>\n",
       "      <td>2017-05-22</td>\n",
       "      <td>Kokkinakis T.</td>\n",
       "      <td>False</td>\n",
       "      <td>2017.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48310</th>\n",
       "      <td>2017-05-30</td>\n",
       "      <td>Kokkinakis T.</td>\n",
       "      <td>False</td>\n",
       "      <td>2017.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49286</th>\n",
       "      <td>2017-09-25</td>\n",
       "      <td>Pavic M.</td>\n",
       "      <td>False</td>\n",
       "      <td>2017.666667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Date         Player  isWinner    DateFloat\n",
       "47124 2017-01-10       Sitak A.     False  2017.000000\n",
       "47225 2017-01-17        Haas T.     False  2017.000000\n",
       "47592 2017-02-28    Alawadhi O.     False  2017.083333\n",
       "48235 2017-05-22  Kokkinakis T.     False  2017.333333\n",
       "48310 2017-05-30  Kokkinakis T.     False  2017.333333\n",
       "49286 2017-09-25       Pavic M.     False  2017.666667"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_nan_Ranks_T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Imputations in progress 0.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "imputeRankData(nanData = df_nan_Ranks_T ,sourceData = df_atp, targetData = df_atp_Test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in WRank column (Test data): 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in WRank column (Test data): \" + str(sum(df_atp_Test.WRank.isna())))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values in LRank column (Test data): 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of missing values in LRank column (Test data): \" + str(sum(df_atp_Test.LRank.isna())))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we have imputed all the missing values with numerical values, we can cast the Rank related variables to 'int64'."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test['WRank'] = df_atp_Test['WRank'].astype('int64')\n",
    "df_atp_Test['LRank'] = df_atp_Test['LRank'].astype('int64')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.3 - Imputation of Point-related missing values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we'll try to handle Point-related missing values. The basic idea is to fit a simple regression model with the expression normalized(Point) = f(log(1+1/Rank)). We won't introduce time variable and deal with the problem as a covariant time series problem to make it simple since it's just about imputations, even if in reality, points notation has changed in the ATP ranking system. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Players having missing points :\n",
    "nanPointsPlayersW = df_atp_Train.loc[df_atp_Train.WPts.isna(),['WRank','Winner']].copy() \n",
    "nanPointsPlayersL = df_atp_Train.loc[df_atp_Train.LPts.isna(),['LRank','Loser']].copy() \n",
    "Trues = np.repeat(True, repeats=nanPointsPlayersW.shape[0], axis=0)\n",
    "Falses = np.repeat(False, repeats=nanPointsPlayersL.shape[0], axis=0)\n",
    "df_nan_Points = pd.DataFrame({'Player':pd.concat([nanPointsPlayersW['Winner'], nanPointsPlayersL['Loser']], axis=0),\n",
    "                                 'Rank':pd.concat([nanPointsPlayersW['WRank'], nanPointsPlayersL['LRank']], axis=0),\n",
    "                                 'isWinner':pd.Series(np.concatenate([Trues,Falses]).flatten(), \n",
    "                                            index=np.concatenate([nanPointsPlayersW.index, nanPointsPlayersL.index]))})\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Preparing data for training the average model Points = f(Rank)\n",
    "cleanData = df_atp_Train.loc[:,['WRank','LRank','WPts','LPts']].copy()\n",
    "cleanData.dropna(axis=0, inplace=True)\n",
    "\n",
    "PtsSerie = pd.concat([cleanData.LPts, cleanData.WPts], axis=0)\n",
    "RankSeries = pd.concat([cleanData.LRank, cleanData.WRank], axis=0)\n",
    "df_rank_points = pd.DataFrame({'Rank':RankSeries,'Points': PtsSerie })\n",
    "\n",
    "df_rank_points.sort_values(axis=0, by='Rank', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Rank</th>\n",
       "      <th>Points</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47027</th>\n",
       "      <td>1</td>\n",
       "      <td>11185.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27108</th>\n",
       "      <td>1</td>\n",
       "      <td>14960.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27049</th>\n",
       "      <td>1</td>\n",
       "      <td>15360.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27047</th>\n",
       "      <td>1</td>\n",
       "      <td>15360.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27039</th>\n",
       "      <td>1</td>\n",
       "      <td>15360.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Rank   Points\n",
       "47027     1  11185.0\n",
       "27108     1  14960.0\n",
       "27049     1  15360.0\n",
       "27047     1  15360.0\n",
       "27039     1  15360.0"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#How rank-points looks like\n",
    "df_rank_points.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's visualize the Rank-Point relationship for a given Date"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_rank_points = df_rank_points.groupby('Rank').aggregate('median')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_rank_points.reset_index(inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Rank</th>\n",
       "      <th>Points</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>12040.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>8700.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>6745.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>5715.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>4840.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Rank   Points\n",
       "0     1  12040.0\n",
       "1     2   8700.0\n",
       "2     3   6745.0\n",
       "3     4   5715.0\n",
       "4     5   4840.0"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_rank_points.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "XRank = df_rank_points.loc[:,'Rank'].values.reshape(-1, 1)\n",
    "YRank = df_rank_points.loc[:,'Points'].values.reshape(-1, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "RfRegressor=RandomForestRegressor(n_estimators=2200, random_state=1995, min_samples_leaf=5, max_features=1, criterion='mse' )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=None,\n",
       "           max_features=1, max_leaf_nodes=None, min_impurity_decrease=0.0,\n",
       "           min_impurity_split=None, min_samples_leaf=5,\n",
       "           min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
       "           n_estimators=2200, n_jobs=1, oob_score=False, random_state=1995,\n",
       "           verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "RfRegressor.fit(XRank, YRank.squeeze())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictedPoints = RfRegressor.predict(XRank)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'True Point values Vs Fitted values')"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10,6))\n",
    "sns.regplot(x= YRank.squeeze(), y = predictedPoints, color='green', ax=ax)\n",
    "plt.title('True Point values Vs Fitted values')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1870.4])"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame({'Point':YRank.squeeze()}).quantile(0.98).values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "More than **98% of Points lay bellow 2000** and thus the model can be used for interpolation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "def avgRankPointPred(ranks ,model = RfRegressor):\n",
    "    \"\"\"\n",
    "    -Overview : \n",
    "    Point Imputations Computing function (based on a pre-trained model passed as argument)\n",
    "    \n",
    "    - Arguments : \n",
    "    ranks : Player's ranks  \n",
    "    model : Fitted model for Prediction\n",
    "    \n",
    "    - Returns :\n",
    "    numpy array of predicted points values by the model passed in argument (will serve for imputation)\n",
    "    \"\"\"\n",
    "    myranks = np.array(ranks).reshape((len(ranks),1))\n",
    "    Ypreds = np.round(model.predict(myranks))\n",
    "    return(Ypreds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computePointImput(Player, nanData = df_nan_Points ,targetData = df_atp_Train):\n",
    "    \"\"\"\n",
    "    - Overview :\n",
    "    Points Imputations Computing function\n",
    "    \n",
    "    - Arguments :\n",
    "    Player : [str] Player's name\n",
    "    nanData : [pandas DataFrame] data of players with Points missing values (plays the role of training data set)\n",
    "    targetData : [pandas DataFrame] target dataset to compute imputations (plays the role of test data set)\n",
    "    \n",
    "    - Returns : \n",
    "    [pandas Data Frame] : subset of nanData for the given Player + computed imputation values of points\n",
    "    \"\"\"\n",
    "    \n",
    "    winData = targetData.loc[targetData.Winner==Player,['WRank', 'WPts']]\n",
    "    loseData = targetData.loc[targetData.Loser==Player,['LRank','LPts']]\n",
    "    \n",
    "    playerNanData = nanData.loc[nanData['Player']==Player,:].copy()\n",
    "    \n",
    "    PointData = pd.DataFrame({ 'Rank':pd.concat([winData['WRank'], loseData['LRank']], axis=0),\n",
    "                             'Points':pd.concat([winData['WPts'], loseData['LPts']], axis=0)})\n",
    "    PointData.sort_index(inplace=True)\n",
    "    PointData.dropna(inplace=True)\n",
    "    predPoints = avgRankPointPred(playerNanData['Rank'])\n",
    "    output = pd.concat([playerNanData, pd.DataFrame({'PredPoint':predPoints}, index=playerNanData.index)], axis=1)\n",
    "    return(output)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "def imputePointsData(nanData = df_nan_Points ,targetData = df_atp_Train):\n",
    "    \"\"\"\n",
    "    - Overview :\n",
    "    Inplace Points Imputation of data passed as targetData based on computed imputations by the previous function 'computePointImput'\n",
    "    \n",
    "    -Arguments :\n",
    "    nanData : [pandas DataFrame] data of players with Points missing values\n",
    "    targetData : [pandas DataFrame] target dataset to be imputed\n",
    "    \n",
    "    -Returns:\n",
    "    Inplace imputations (no return)\n",
    "    Prints imputations progress\n",
    "    \"\"\"\n",
    "    progress = 0\n",
    "    nanPtsLen = len(nanData['Player'].unique())\n",
    "    tenPer, per = nanPtsLen//10, 0\n",
    "    for p in nanData['Player'].unique():\n",
    "        imputations = computePointImput(Player=p, nanData = nanData ,targetData = targetData)        \n",
    "        for i,val in enumerate(imputations['isWinner']):\n",
    "            imp = imputations.loc[imputations.index[i], 'PredPoint']\n",
    "            if (val): #Winner case\n",
    "                targetData.loc[imputations.index[i],'WPts'] = imp\n",
    "            else:     #Loser case\n",
    "                targetData.loc[imputations.index[i],'LPts'] = imp\n",
    "        if(per==progress):\n",
    "            print(\"Imputations in progress \"+str( np.round(100*per/nanPtsLen, decimals=0))+\"%\")\n",
    "            per += tenPer\n",
    "        progress += 1\n",
    "        if(progress==nanPtsLen):\n",
    "            print(\"Done !\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Imputations in progress 0.0%\n",
      "Imputations in progress 10.0%\n",
      "Imputations in progress 20.0%\n",
      "Imputations in progress 30.0%\n",
      "Imputations in progress 40.0%\n",
      "Imputations in progress 50.0%\n",
      "Imputations in progress 60.0%\n",
      "Imputations in progress 69.0%\n",
      "Imputations in progress 79.0%\n",
      "Imputations in progress 89.0%\n",
      "Imputations in progress 99.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "imputePointsData(nanData = df_nan_Points ,targetData = df_atp_Train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percentage of missing values in WPts column : 0.0%\n",
      "Percentage of missing values in LPts column : 0.0%\n"
     ]
    }
   ],
   "source": [
    "print(\"Percentage of missing values in WPts column : \" + str(np.round(100*sum(df_atp_Train.WPts.isna())/len(df_atp_Train), 3))+\"%\")\n",
    "print(\"Percentage of missing values in LPts column : \" + str(np.round(100*sum(df_atp_Train.LPts.isna())/len(df_atp_Train), 3))+\"%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train['WPts'] = df_atp_Train['WPts'].astype('int64')\n",
    "df_atp_Train['LPts'] = df_atp_Train['LPts'].astype('int64')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Quick comparison between some observations before/after na imputation:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>755</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>5858</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>1025</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>736</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>625</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament  WPts  WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   755     63   \n",
       "1  International    Hard  Australian Hardcourt Championships  5858      5   \n",
       "2  International    Hard  Australian Hardcourt Championships  1025     40   \n",
       "3  International    Hard  Australian Hardcourt Championships   736     65   \n",
       "4  International    Hard  Australian Hardcourt Championships   625     81   \n",
       "\n",
       "        Winner  \n",
       "0   Dosedel S.  \n",
       "1   Enqvist T.  \n",
       "2    Escude N.  \n",
       "3   Federer R.  \n",
       "4  Fromberg R.  "
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>NaN</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>NaN</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>NaN</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>NaN</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>NaN</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>NaN</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>NaN</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>NaN</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>NaN</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>NaN</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   NaN    77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   NaN    56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03   NaN   655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   NaN    87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   NaN   198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament  WPts WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   NaN    63   \n",
       "1  International    Hard  Australian Hardcourt Championships   NaN     5   \n",
       "2  International    Hard  Australian Hardcourt Championships   NaN    40   \n",
       "3  International    Hard  Australian Hardcourt Championships   NaN    65   \n",
       "4  International    Hard  Australian Hardcourt Championships   NaN    81   \n",
       "\n",
       "        Winner  \n",
       "0   Dosedel S.  \n",
       "1   Enqvist T.  \n",
       "2    Escude N.  \n",
       "3   Federer R.  \n",
       "4  Fromberg R.  "
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp.loc[:,usedColumns].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we do the same for the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percentage of missing values in WPts column (Test) : 0.0%\n",
      "Percentage of missing values in LPts column (Test) : 0.228%\n"
     ]
    }
   ],
   "source": [
    "print(\"Percentage of missing values in WPts column (Test) : \" + str(np.round(100*sum(df_atp_Test.WPts.isna())/len(df_atp_Test), 3))+\"%\")\n",
    "print(\"Percentage of missing values in LPts column (Test) : \" + str(np.round(100*sum(df_atp_Test.LPts.isna())/len(df_atp_Test), 3))+\"%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Players having missing points :\n",
    "nanPointsPlayersL_T = df_atp_Test.loc[df_atp_Test.LPts.isna(),['LRank','Loser']].copy() \n",
    "FalsesT = np.repeat(False, repeats=nanPointsPlayersL_T.shape[0], axis=0)\n",
    "df_nan_Points_Test = pd.DataFrame({'Player':nanPointsPlayersL_T['Loser'],\n",
    "                                 'Rank':nanPointsPlayersL_T['LRank'],\n",
    "                                 'isWinner':pd.Series(FalsesT, index = nanPointsPlayersL_T.index)\n",
    "                                  })\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Imputations in progress 0.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "imputePointsData(nanData = df_nan_Points_Test ,targetData = df_atp_Test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percentage of missing values in WPts column (Test) : 0.0%\n",
      "Percentage of missing values in LPts column (Test) : 0.0%\n"
     ]
    }
   ],
   "source": [
    "print(\"Percentage of missing values in WPts column (Test) : \" + str(np.round(100*sum(df_atp_Test.WPts.isna())/len(df_atp_Test), 3))+\"%\")\n",
    "print(\"Percentage of missing values in LPts column (Test) : \" + str(np.round(100*sum(df_atp_Test.LPts.isna())/len(df_atp_Test), 3))+\"%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test['WPts'] = df_atp_Test['WPts'].astype('int64')\n",
    "df_atp_Test['LPts'] = df_atp_Test['LPts'].astype('int64')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Engineering  \n",
    "### Output modeling  \n",
    "If we define the outcome of an ATP match as the 'Winner's name', the problem would be a multi-classification task with an output taking labels belonging to the names of all players since 2000. This 'naive' approach is not realistic since for each match we have two players and the winner is among them. One way to explore that is to consider two aditionnal features 'player1' and 'player2' to which we randomly assign the match winner and loser, the target variable would belong to the couple (0,1) and thus the problem can be treated as a **binary classification** learning."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>755</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>5858</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>Clement A.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>1025</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>Baccanello P.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>736</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>625</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament  WPts  WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   755     63   \n",
       "1  International    Hard  Australian Hardcourt Championships  5858      5   \n",
       "2  International    Hard  Australian Hardcourt Championships  1025     40   \n",
       "3  International    Hard  Australian Hardcourt Championships   736     65   \n",
       "4  International    Hard  Australian Hardcourt Championships   625     81   \n",
       "\n",
       "        Winner      Player1         Player2  \n",
       "0   Dosedel S.   Dosedel S.     Ljubicic I.  \n",
       "1   Enqvist T.   Enqvist T.      Clement A.  \n",
       "2    Escude N.    Escude N.   Baccanello P.  \n",
       "3   Federer R.   Federer R.  Knippschild J.  \n",
       "4  Fromberg R.  Fromberg R.   Woodbridge T.  "
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train[['Player1','Player2']] = df_atp_Train.loc[:,['Winner','Loser']]\n",
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(1995)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "Won = np.random.randint(2,size=len(df_atp_Train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train['Won'] = Won\n",
    "boolWon = (Won==0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train.loc[boolWon,'Player1'] , df_atp_Train.loc[boolWon,'Player2'] = df_atp_Train.loc[boolWon,'Player2'], df_atp_Train.loc[boolWon,'Player1'] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>755</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>5858</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>1025</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>736</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>625</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament  WPts  WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   755     63   \n",
       "1  International    Hard  Australian Hardcourt Championships  5858      5   \n",
       "2  International    Hard  Australian Hardcourt Championships  1025     40   \n",
       "3  International    Hard  Australian Hardcourt Championships   736     65   \n",
       "4  International    Hard  Australian Hardcourt Championships   625     81   \n",
       "\n",
       "        Winner        Player1         Player2  Won  \n",
       "0   Dosedel S.    Ljubicic I.      Dosedel S.    0  \n",
       "1   Enqvist T.     Clement A.      Enqvist T.    0  \n",
       "2    Escude N.  Baccanello P.       Escude N.    0  \n",
       "3   Federer R.     Federer R.  Knippschild J.    1  \n",
       "4  Fromberg R.    Fromberg R.   Woodbridge T.    1  "
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We do the same for the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47028</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-01</td>\n",
       "      <td>372</td>\n",
       "      <td>160</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>689</td>\n",
       "      <td>79</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Ymer E.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47029</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>854</td>\n",
       "      <td>54</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1102</td>\n",
       "      <td>39</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Robert S.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47030</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1465</td>\n",
       "      <td>26</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1785</td>\n",
       "      <td>21</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Tomic B.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47031</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>443</td>\n",
       "      <td>141</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1001</td>\n",
       "      <td>45</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Escobedo E.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47032</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1320</td>\n",
       "      <td>33</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>2035</td>\n",
       "      <td>17</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Johnson S.</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Best of    Court       Date  LPts  LRank        Loser      Round  \\\n",
       "47028        3  Outdoor 2017-01-01   372    160      Ymer E.  1st Round   \n",
       "47029        3  Outdoor 2017-01-02   854     54    Robert S.  1st Round   \n",
       "47030        3  Outdoor 2017-01-02  1465     26     Tomic B.  1st Round   \n",
       "47031        3  Outdoor 2017-01-02   443    141  Escobedo E.  1st Round   \n",
       "47032        3  Outdoor 2017-01-02  1320     33   Johnson S.  1st Round   \n",
       "\n",
       "       Series Surface              Tournament  WPts  WRank       Winner  \\\n",
       "47028  ATP250    Hard  Brisbane International   689     79  Thompson J.   \n",
       "47029  ATP250    Hard  Brisbane International  1102     39     Mahut N.   \n",
       "47030  ATP250    Hard  Brisbane International  1785     21    Ferrer D.   \n",
       "47031  ATP250    Hard  Brisbane International  1001     45    Edmund K.   \n",
       "47032  ATP250    Hard  Brisbane International  2035     17  Dimitrov G.   \n",
       "\n",
       "           Player1      Player2  \n",
       "47028  Thompson J.      Ymer E.  \n",
       "47029     Mahut N.    Robert S.  \n",
       "47030    Ferrer D.     Tomic B.  \n",
       "47031    Edmund K.  Escobedo E.  \n",
       "47032  Dimitrov G.   Johnson S.  "
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Test[['Player1','Player2']] = df_atp_Test.loc[:,['Winner','Loser']]\n",
    "df_atp_Test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [],
   "source": [
    "WonT = np.random.randint(2,size=len(df_atp_Test))\n",
    "boolWonT = (WonT==0)\n",
    "df_atp_Test['Won'] = WonT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47028</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-01</td>\n",
       "      <td>372</td>\n",
       "      <td>160</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>689</td>\n",
       "      <td>79</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47029</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>854</td>\n",
       "      <td>54</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1102</td>\n",
       "      <td>39</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47030</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1465</td>\n",
       "      <td>26</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1785</td>\n",
       "      <td>21</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47031</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>443</td>\n",
       "      <td>141</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1001</td>\n",
       "      <td>45</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47032</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1320</td>\n",
       "      <td>33</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>2035</td>\n",
       "      <td>17</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Best of    Court       Date  LPts  LRank        Loser      Round  \\\n",
       "47028        3  Outdoor 2017-01-01   372    160      Ymer E.  1st Round   \n",
       "47029        3  Outdoor 2017-01-02   854     54    Robert S.  1st Round   \n",
       "47030        3  Outdoor 2017-01-02  1465     26     Tomic B.  1st Round   \n",
       "47031        3  Outdoor 2017-01-02   443    141  Escobedo E.  1st Round   \n",
       "47032        3  Outdoor 2017-01-02  1320     33   Johnson S.  1st Round   \n",
       "\n",
       "       Series Surface              Tournament  WPts  WRank       Winner  \\\n",
       "47028  ATP250    Hard  Brisbane International   689     79  Thompson J.   \n",
       "47029  ATP250    Hard  Brisbane International  1102     39     Mahut N.   \n",
       "47030  ATP250    Hard  Brisbane International  1785     21    Ferrer D.   \n",
       "47031  ATP250    Hard  Brisbane International  1001     45    Edmund K.   \n",
       "47032  ATP250    Hard  Brisbane International  2035     17  Dimitrov G.   \n",
       "\n",
       "           Player1      Player2  Won  \n",
       "47028  Thompson J.      Ymer E.    0  \n",
       "47029     Mahut N.    Robert S.    0  \n",
       "47030    Ferrer D.     Tomic B.    1  \n",
       "47031    Edmund K.  Escobedo E.    0  \n",
       "47032  Dimitrov G.   Johnson S.    0  "
      ]
     },
     "execution_count": 119,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test.loc[boolWonT,'Player1'] , df_atp_Test.loc[boolWonT,'Player2'] = df_atp_Test.loc[boolWonT,'Player2'], df_atp_Test.loc[boolWonT,'Player1'] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47028</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-01</td>\n",
       "      <td>372</td>\n",
       "      <td>160</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>689</td>\n",
       "      <td>79</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47029</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>854</td>\n",
       "      <td>54</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1102</td>\n",
       "      <td>39</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47030</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1465</td>\n",
       "      <td>26</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1785</td>\n",
       "      <td>21</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47031</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>443</td>\n",
       "      <td>141</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1001</td>\n",
       "      <td>45</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47032</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1320</td>\n",
       "      <td>33</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>2035</td>\n",
       "      <td>17</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Best of    Court       Date  LPts  LRank        Loser      Round  \\\n",
       "47028        3  Outdoor 2017-01-01   372    160      Ymer E.  1st Round   \n",
       "47029        3  Outdoor 2017-01-02   854     54    Robert S.  1st Round   \n",
       "47030        3  Outdoor 2017-01-02  1465     26     Tomic B.  1st Round   \n",
       "47031        3  Outdoor 2017-01-02   443    141  Escobedo E.  1st Round   \n",
       "47032        3  Outdoor 2017-01-02  1320     33   Johnson S.  1st Round   \n",
       "\n",
       "       Series Surface              Tournament  WPts  WRank       Winner  \\\n",
       "47028  ATP250    Hard  Brisbane International   689     79  Thompson J.   \n",
       "47029  ATP250    Hard  Brisbane International  1102     39     Mahut N.   \n",
       "47030  ATP250    Hard  Brisbane International  1785     21    Ferrer D.   \n",
       "47031  ATP250    Hard  Brisbane International  1001     45    Edmund K.   \n",
       "47032  ATP250    Hard  Brisbane International  2035     17  Dimitrov G.   \n",
       "\n",
       "           Player1      Player2  Won  \n",
       "47028      Ymer E.  Thompson J.    0  \n",
       "47029    Robert S.     Mahut N.    0  \n",
       "47030    Ferrer D.     Tomic B.    1  \n",
       "47031  Escobedo E.    Edmund K.    0  \n",
       "47032   Johnson S.  Dimitrov G.    0  "
      ]
     },
     "execution_count": 121,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Test.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature creation  \n",
    "#### Player1 Vs Player2  \n",
    "head-to-head wins variable : count of all times when player1 beated player2 minus the count of all times when player2 beated player1 before the year of the match (since it doesn't make sense to aggregate with the whole data, only past records should explain future results).  \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train['Year'] = df_atp_Train.Date.apply(lambda x : x.year)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test['Year'] = df_atp_Test.Date.apply(lambda x : x.year)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [],
   "source": [
    "def recordedVs(Year, data, past = 17, includeYear = False):\n",
    "    \"\"\"\n",
    "    Overview:\n",
    "    For a given year, computes all the recorded relative victories of the duels. (Nb of winnings/ Nb of matches)\n",
    "    \n",
    "    Arguments:\n",
    "    - Year : [scalar] the year of the match, the computations will use the data recorded before that year\n",
    "    - data : [data frame] Train data set that contains the whole records from 2000 to 2016\n",
    "    - past : number of years to be used for \n",
    "    \n",
    "    Returns : \n",
    "    A dataframe with the Winner, Loser, winning rate of the winner in the past (<Year)\n",
    "    \n",
    "    *The victories are also weighted with the number of Sets played (Best Of) since a value of 3 means at least two\n",
    "    successive victories (high margin) while 5 indicates a certain difficulty (low margin )\n",
    "    \n",
    "    \n",
    "    \"\"\"\n",
    "    pastYears = np.min([past, Year-2000])   #Since the oldest record was in year 2000\n",
    "    if includeYear:\n",
    "        includeY = 1\n",
    "    else:\n",
    "        includeY = 0\n",
    "    \n",
    "    df_sub = data.loc[(Year-pastYears <= data.Year) & (data.Year<Year+includeY),['Best of','Winner','Loser']].copy()\n",
    "    df_sub['CouplesSet'] = df_sub.apply(lambda x : str(set((x['Winner'], x['Loser']))), axis=1)\n",
    "    countSets = df_sub.CouplesSet.value_counts()\n",
    "    countSets = pd.DataFrame({'couplePlayer':countSets.index, 'NMatche':countSets.values})\n",
    "    countWons = df_sub.groupby(['Winner','Loser'])\n",
    "    countWons = countWons.aggregate(['count','mean'])\n",
    "    countWons.reset_index(inplace=True)\n",
    "    countWons['CountWon'] = countWons.loc[:,('Best of','count')]\n",
    "    countWons['MeanBestOf'] = countWons.loc[:,('Best of','mean')]\n",
    "    countWons.drop(axis=1, columns='Best of', level=0, inplace=True)\n",
    "    countWons['relativeVictory'] = 0.0\n",
    "    countWIndx = countWons.index\n",
    "    countSIndx = countSets.index\n",
    "    \n",
    "    for i in range(len(countWons)):\n",
    "        p1 = countWons.loc[countWIndx[i],'Winner'].values[0]\n",
    "        p2 = countWons.loc[countWIndx[i],'Loser'].values[0]\n",
    "    \n",
    "        Nindex = countSIndx[countSets.couplePlayer.apply(lambda x : (p1 in x) & (p2 in x) )]\n",
    "        N = countSets.loc[Nindex, 'NMatche']\n",
    "        countWons.loc[countWIndx[i],'relativeVictory'] = np.asscalar(countWons.loc[countWIndx[i],'CountWon']/N.values[0])\n",
    "    return(countWons)\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [],
   "source": [
    "genHistory = recordedVs(2017, data = df_atp_Train, past = 17, includeYear=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [],
   "source": [
    "genHistoryTest = recordedVs(2017, data = df_atp_Train, past = 10, includeYear=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here are some match cases with the computed values of head to head results. For example, Federer won Agassi in 80% of their games, so he lost the left 20%. We see it clearly in the observations bellow."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>Winner</th>\n",
       "      <th>Loser</th>\n",
       "      <th>CountWon</th>\n",
       "      <th>MeanBestOf</th>\n",
       "      <th>relativeVictory</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>7913</th>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Agassi A.</td>\n",
       "      <td>8</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Winner      Loser CountWon MeanBestOf relativeVictory\n",
       "                                                               \n",
       "7913  Federer R.  Agassi A.        8        4.0             0.8"
      ]
     },
     "execution_count": 127,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "genHistory.loc[(genHistory.Winner== 'Federer R.')&(genHistory.Loser=='Agassi A.'),:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>Winner</th>\n",
       "      <th>Loser</th>\n",
       "      <th>CountWon</th>\n",
       "      <th>MeanBestOf</th>\n",
       "      <th>relativeVictory</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>151</th>\n",
       "      <td>Agassi A.</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>2</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Winner       Loser CountWon MeanBestOf relativeVictory\n",
       "                                                              \n",
       "151  Agassi A.  Federer R.        2        5.0             0.2"
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "genHistory.loc[(genHistory.Winner == 'Agassi A.') &(genHistory.Loser=='Federer R.'),:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Another special case is when a player X have won all the matches against player Y, so the results will be 100% and 0%."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>Winner</th>\n",
       "      <th>Loser</th>\n",
       "      <th>CountWon</th>\n",
       "      <th>MeanBestOf</th>\n",
       "      <th>relativeVictory</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11954</th>\n",
       "      <td>Hajek J.</td>\n",
       "      <td>Ulihrach B.</td>\n",
       "      <td>1</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         Winner        Loser CountWon MeanBestOf relativeVictory\n",
       "                                                                \n",
       "11954  Hajek J.  Ulihrach B.        1        5.0             1.0"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "genHistory.loc[(genHistory.Winner=='Hajek J.') & (genHistory.Loser=='Ulihrach B.'),:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>Winner</th>\n",
       "      <th>Loser</th>\n",
       "      <th>CountWon</th>\n",
       "      <th>MeanBestOf</th>\n",
       "      <th>relativeVictory</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [(Winner, ), (Loser, ), (CountWon, ), (MeanBestOf, ), (relativeVictory, )]\n",
       "Index: []"
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "genHistory.loc[(genHistory.Loser=='Hajek J.') & (genHistory.Winner=='Ulihrach B.'),:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can fill the values for the new features 'p1Vsp2' and 'Wp1Vsp2' based on the records in the 'genHistory' dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train['p1Vsp2'] = 0\n",
    "df_atp_Train['Wp1Vsp2'] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [],
   "source": [
    "def p1Vsp2(data, history = genHistory):\n",
    "\n",
    "    dataIndex = data.index\n",
    "    progress = 0\n",
    "    dataLenght = len(data)\n",
    "    tenPer, per = dataLenght//10, 0\n",
    "\n",
    "    for i in dataIndex:\n",
    "        p1 = data.loc[i,'Player1']\n",
    "        p2 = data.loc[i,'Player2']\n",
    "        relVict = history.loc[((history.Winner==p1) & (history.Loser==p2)),'relativeVictory' ]\n",
    "        relLoss = history.loc[((history.Winner==p2) & (history.Loser==p1)),'relativeVictory' ]\n",
    "        \n",
    "        if len(relVict)==0:\n",
    "            relVict = 0\n",
    "        else:\n",
    "            relVict = np.asscalar(relVict.values)\n",
    "        if len(relLoss)==0:\n",
    "            relLoss = 0  \n",
    "        else:\n",
    "            relLoss = np.asscalar(relLoss.values)\n",
    "        \n",
    "     \n",
    "        MeanSetsW = history.loc[((history.Winner==p1) & (history.Loser==p2)),'MeanBestOf' ]\n",
    "        MeanSetsL = history.loc[((history.Loser==p1) & (history.Winner==p2)),'MeanBestOf' ]\n",
    "    \n",
    "        if len(MeanSetsW)==0:\n",
    "            MeanSetsW = 3\n",
    "        else:\n",
    "            MeanSetsW = np.asscalar(MeanSetsW.values)\n",
    "        if len(MeanSetsL)==0:\n",
    "            MeanSetsL = 3\n",
    "        else:\n",
    "            MeanSetsL = np.asscalar(MeanSetsL.values)\n",
    "        \n",
    "        WrelVict = relVict/MeanSetsW\n",
    "        LrelVict = relLoss/MeanSetsL\n",
    "    \n",
    "    \n",
    "        data.loc[i,'p1Vsp2'] = relVict - relLoss\n",
    "        data.loc[i,'Wp1Vsp2'] = WrelVict - LrelVict\n",
    "    \n",
    "        if(per==progress):\n",
    "            print(\"Computations in progress \"+str( np.round(100*per/dataLenght, decimals=0))+\"%\")\n",
    "            per += tenPer\n",
    "        progress += 1\n",
    "        if(progress==dataLenght):\n",
    "            print(\"Done !\")\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computations in progress 0.0%\n",
      "Computations in progress 10.0%\n",
      "Computations in progress 20.0%\n",
      "Computations in progress 30.0%\n",
      "Computations in progress 40.0%\n",
      "Computations in progress 50.0%\n",
      "Computations in progress 60.0%\n",
      "Computations in progress 70.0%\n",
      "Computations in progress 80.0%\n",
      "Computations in progress 90.0%\n",
      "Computations in progress 100.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "p1Vsp2(data = df_atp_Train, history = genHistory)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>755</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>5858</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-0.5</td>\n",
       "      <td>-0.200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>1025</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>736</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>625</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament  WPts  WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   755     63   \n",
       "1  International    Hard  Australian Hardcourt Championships  5858      5   \n",
       "2  International    Hard  Australian Hardcourt Championships  1025     40   \n",
       "3  International    Hard  Australian Hardcourt Championships   736     65   \n",
       "4  International    Hard  Australian Hardcourt Championships   625     81   \n",
       "\n",
       "        Winner        Player1         Player2  Won  Year  p1Vsp2   Wp1Vsp2  \n",
       "0   Dosedel S.    Ljubicic I.      Dosedel S.    0  2000    -1.0 -0.333333  \n",
       "1   Enqvist T.     Clement A.      Enqvist T.    0  2000    -0.5 -0.200000  \n",
       "2    Escude N.  Baccanello P.       Escude N.    0  2000    -1.0 -0.333333  \n",
       "3   Federer R.     Federer R.  Knippschild J.    1  2000     1.0  0.333333  \n",
       "4  Fromberg R.    Fromberg R.   Woodbridge T.    1  2000     1.0  0.333333  "
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's visualize the relationship between these added features and the output 'Won'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Boxplot of p1Vsp2 varialbe by winning case (Train Set)')"
      ]
     },
     "execution_count": 135,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10,6))\n",
    "bplot = sns.boxplot(y='p1Vsp2', x='Won', data=df_atp_Train, palette=\"colorblind\")\n",
    "plt.title('Boxplot of p1Vsp2 varialbe by winning case (Train Set)')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Boxplot of Wp1Vsp2 varialbe by winning case (Train Set)')"
      ]
     },
     "execution_count": 136,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10,6))\n",
    "bplot = sns.boxplot(y='Wp1Vsp2', x='Won', data=df_atp_Train, palette=\"colorblind\")\n",
    "plt.title('Boxplot of Wp1Vsp2 varialbe by winning case (Train Set)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have clearly a great separation between the winning/losing cases from the angle of the two added features 'p1Vsp2' and 'p2Vsp1'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We create the same 'Head to Head' features for the Test set based on recent historical data (from 2012->2017)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test['p1Vsp2'] = 0\n",
    "df_atp_Test['Wp1Vsp2'] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computations in progress 0.0%\n",
      "Computations in progress 10.0%\n",
      "Computations in progress 20.0%\n",
      "Computations in progress 30.0%\n",
      "Computations in progress 40.0%\n",
      "Computations in progress 50.0%\n",
      "Computations in progress 60.0%\n",
      "Computations in progress 70.0%\n",
      "Computations in progress 80.0%\n",
      "Computations in progress 90.0%\n",
      "Computations in progress 100.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "p1Vsp2(data = df_atp_Test, history = genHistoryTest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>WPts</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47028</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-01</td>\n",
       "      <td>372</td>\n",
       "      <td>160</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>689</td>\n",
       "      <td>79</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47029</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>854</td>\n",
       "      <td>54</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1102</td>\n",
       "      <td>39</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47030</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1465</td>\n",
       "      <td>26</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1785</td>\n",
       "      <td>21</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1</td>\n",
       "      <td>2017</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.079365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47031</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>443</td>\n",
       "      <td>141</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>1001</td>\n",
       "      <td>45</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47032</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1320</td>\n",
       "      <td>33</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>2035</td>\n",
       "      <td>17</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-0.106061</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Best of    Court       Date  LPts  LRank        Loser      Round  \\\n",
       "47028        3  Outdoor 2017-01-01   372    160      Ymer E.  1st Round   \n",
       "47029        3  Outdoor 2017-01-02   854     54    Robert S.  1st Round   \n",
       "47030        3  Outdoor 2017-01-02  1465     26     Tomic B.  1st Round   \n",
       "47031        3  Outdoor 2017-01-02   443    141  Escobedo E.  1st Round   \n",
       "47032        3  Outdoor 2017-01-02  1320     33   Johnson S.  1st Round   \n",
       "\n",
       "       Series Surface              Tournament  WPts  WRank       Winner  \\\n",
       "47028  ATP250    Hard  Brisbane International   689     79  Thompson J.   \n",
       "47029  ATP250    Hard  Brisbane International  1102     39     Mahut N.   \n",
       "47030  ATP250    Hard  Brisbane International  1785     21    Ferrer D.   \n",
       "47031  ATP250    Hard  Brisbane International  1001     45    Edmund K.   \n",
       "47032  ATP250    Hard  Brisbane International  2035     17  Dimitrov G.   \n",
       "\n",
       "           Player1      Player2  Won  Year    p1Vsp2   Wp1Vsp2  \n",
       "47028      Ymer E.  Thompson J.    0  2017  0.000000  0.000000  \n",
       "47029    Robert S.     Mahut N.    0  2017  0.000000  0.000000  \n",
       "47030    Ferrer D.     Tomic B.    1  2017  0.333333  0.079365  \n",
       "47031  Escobedo E.    Edmund K.    0  2017 -1.000000 -0.200000  \n",
       "47032   Johnson S.  Dimitrov G.    0  2017 -0.333333 -0.106061  "
      ]
     },
     "execution_count": 139,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Boxplot of p1Vsp2 varialbe by winning case (Test Set)')"
      ]
     },
     "execution_count": 140,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10,6))\n",
    "bplot = sns.boxplot(y='p1Vsp2', x='Won', data=df_atp_Test, palette=\"colorblind\")\n",
    "plt.title('Boxplot of p1Vsp2 varialbe by winning case (Test Set)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Separation seems more challenging in the Test set, this will probably lead to some misclassifications in the model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Rank and Points differences (Player1 - Player2)    \n",
    "The two features we're going to introduce are simply the difference between players performances (Points and Ranks).  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train[\"RankDiff\"] = 0\n",
    "df_atp_Train[\"PtsDiff\"] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [],
   "source": [
    "def performanceDiff(variable = 'Rank', targetData=df_atp_Train):\n",
    "    \"\"\"\n",
    "    Overview:\n",
    "    Computing and filling the difference of a given variable between the two players p1-p2\n",
    "    \n",
    "    Arguments : \n",
    "    variable : 'Rank' or 'Point'\n",
    "    targetData : dataframe to be filled with differences (must contain at least the )\n",
    "    \n",
    "    Returns :\n",
    "    inplace filling of the differences\n",
    "\n",
    "    \"\"\"\n",
    "    if variable == 'Rank':\n",
    "        Wval, Lval = 'WRank', 'LRank'\n",
    "        target = \"RankDiff\"\n",
    "        differences = targetData[Lval]-targetData[Wval]\n",
    "    elif variable == 'Point':\n",
    "        Wval, Lval = 'WPts', 'LPts'\n",
    "        target = \"PtsDiff\"\n",
    "        differences = targetData[Wval]-targetData[Lval]\n",
    "    \n",
    "    targetData[target] = differences\n",
    "    wonInd1 = targetData['Won'] == 0   #The cases where p1=Loser and p2=Winner \n",
    "    targetData.loc[wonInd1,target] = targetData.loc[wonInd1,target].apply(lambda x : -x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [],
   "source": [
    "performanceDiff(variable='Rank', targetData=df_atp_Train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>...</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-0.5</td>\n",
       "      <td>-0.200000</td>\n",
       "      <td>-51</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-615</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>22</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>117</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament   ...     WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   ...        63   \n",
       "1  International    Hard  Australian Hardcourt Championships   ...         5   \n",
       "2  International    Hard  Australian Hardcourt Championships   ...        40   \n",
       "3  International    Hard  Australian Hardcourt Championships   ...        65   \n",
       "4  International    Hard  Australian Hardcourt Championships   ...        81   \n",
       "\n",
       "        Winner        Player1         Player2 Won  Year  p1Vsp2   Wp1Vsp2  \\\n",
       "0   Dosedel S.    Ljubicic I.      Dosedel S.   0  2000    -1.0 -0.333333   \n",
       "1   Enqvist T.     Clement A.      Enqvist T.   0  2000    -0.5 -0.200000   \n",
       "2    Escude N.  Baccanello P.       Escude N.   0  2000    -1.0 -0.333333   \n",
       "3   Federer R.     Federer R.  Knippschild J.   1  2000     1.0  0.333333   \n",
       "4  Fromberg R.    Fromberg R.   Woodbridge T.   1  2000     1.0  0.333333   \n",
       "\n",
       "   RankDiff  PtsDiff  \n",
       "0       -14        0  \n",
       "1       -51        0  \n",
       "2      -615        0  \n",
       "3        22        0  \n",
       "4       117        0  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 144,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "performanceDiff(variable='Point', targetData=df_atp_Train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>...</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>63</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-14</td>\n",
       "      <td>-107</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-0.5</td>\n",
       "      <td>-0.200000</td>\n",
       "      <td>-51</td>\n",
       "      <td>-5040</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>40</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-615</td>\n",
       "      <td>-989</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>65</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>22</td>\n",
       "      <td>148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>81</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>117</td>\n",
       "      <td>371</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament   ...     WRank  \\\n",
       "0  International    Hard  Australian Hardcourt Championships   ...        63   \n",
       "1  International    Hard  Australian Hardcourt Championships   ...         5   \n",
       "2  International    Hard  Australian Hardcourt Championships   ...        40   \n",
       "3  International    Hard  Australian Hardcourt Championships   ...        65   \n",
       "4  International    Hard  Australian Hardcourt Championships   ...        81   \n",
       "\n",
       "        Winner        Player1         Player2 Won  Year  p1Vsp2   Wp1Vsp2  \\\n",
       "0   Dosedel S.    Ljubicic I.      Dosedel S.   0  2000    -1.0 -0.333333   \n",
       "1   Enqvist T.     Clement A.      Enqvist T.   0  2000    -0.5 -0.200000   \n",
       "2    Escude N.  Baccanello P.       Escude N.   0  2000    -1.0 -0.333333   \n",
       "3   Federer R.     Federer R.  Knippschild J.   1  2000     1.0  0.333333   \n",
       "4  Fromberg R.    Fromberg R.   Woodbridge T.   1  2000     1.0  0.333333   \n",
       "\n",
       "   RankDiff  PtsDiff  \n",
       "0       -14     -107  \n",
       "1       -51    -5040  \n",
       "2      -615     -989  \n",
       "3        22      148  \n",
       "4       117      371  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 146,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We do the same thing for the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test[\"RankDiff\"] = 0\n",
    "df_atp_Test[\"PtsDiff\"] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [],
   "source": [
    "performanceDiff(variable='Rank', targetData=df_atp_Test)\n",
    "performanceDiff(variable='Point', targetData=df_atp_Test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>...</th>\n",
       "      <th>WRank</th>\n",
       "      <th>Winner</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47028</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-01</td>\n",
       "      <td>372</td>\n",
       "      <td>160</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>...</td>\n",
       "      <td>79</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>Ymer E.</td>\n",
       "      <td>Thompson J.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-81</td>\n",
       "      <td>-317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47029</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>854</td>\n",
       "      <td>54</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>...</td>\n",
       "      <td>39</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>Robert S.</td>\n",
       "      <td>Mahut N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-15</td>\n",
       "      <td>-248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47030</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1465</td>\n",
       "      <td>26</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>...</td>\n",
       "      <td>21</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Ferrer D.</td>\n",
       "      <td>Tomic B.</td>\n",
       "      <td>1</td>\n",
       "      <td>2017</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.079365</td>\n",
       "      <td>5</td>\n",
       "      <td>320</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47031</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>443</td>\n",
       "      <td>141</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>...</td>\n",
       "      <td>45</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>Escobedo E.</td>\n",
       "      <td>Edmund K.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>-0.200000</td>\n",
       "      <td>-96</td>\n",
       "      <td>-558</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47032</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2017-01-02</td>\n",
       "      <td>1320</td>\n",
       "      <td>33</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>ATP250</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Brisbane International</td>\n",
       "      <td>...</td>\n",
       "      <td>17</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>Johnson S.</td>\n",
       "      <td>Dimitrov G.</td>\n",
       "      <td>0</td>\n",
       "      <td>2017</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-0.106061</td>\n",
       "      <td>-16</td>\n",
       "      <td>-715</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       Best of    Court       Date  LPts  LRank        Loser      Round  \\\n",
       "47028        3  Outdoor 2017-01-01   372    160      Ymer E.  1st Round   \n",
       "47029        3  Outdoor 2017-01-02   854     54    Robert S.  1st Round   \n",
       "47030        3  Outdoor 2017-01-02  1465     26     Tomic B.  1st Round   \n",
       "47031        3  Outdoor 2017-01-02   443    141  Escobedo E.  1st Round   \n",
       "47032        3  Outdoor 2017-01-02  1320     33   Johnson S.  1st Round   \n",
       "\n",
       "       Series Surface              Tournament   ...     WRank       Winner  \\\n",
       "47028  ATP250    Hard  Brisbane International   ...        79  Thompson J.   \n",
       "47029  ATP250    Hard  Brisbane International   ...        39     Mahut N.   \n",
       "47030  ATP250    Hard  Brisbane International   ...        21    Ferrer D.   \n",
       "47031  ATP250    Hard  Brisbane International   ...        45    Edmund K.   \n",
       "47032  ATP250    Hard  Brisbane International   ...        17  Dimitrov G.   \n",
       "\n",
       "           Player1      Player2 Won  Year    p1Vsp2   Wp1Vsp2  RankDiff  \\\n",
       "47028      Ymer E.  Thompson J.   0  2017  0.000000  0.000000       -81   \n",
       "47029    Robert S.     Mahut N.   0  2017  0.000000  0.000000       -15   \n",
       "47030    Ferrer D.     Tomic B.   1  2017  0.333333  0.079365         5   \n",
       "47031  Escobedo E.    Edmund K.   0  2017 -1.000000 -0.200000       -96   \n",
       "47032   Johnson S.  Dimitrov G.   0  2017 -0.333333 -0.106061       -16   \n",
       "\n",
       "       PtsDiff  \n",
       "47028     -317  \n",
       "47029     -248  \n",
       "47030      320  \n",
       "47031     -558  \n",
       "47032     -715  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Test.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Per Round type winning percentage  \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [],
   "source": [
    "uniquePlayers =df_atp_Train.loc[:,['Winner','Loser']].unstack().unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_Unique_Players = pd.DataFrame({'Player':uniquePlayers,'FirstRounds':0, \n",
    "                                  'Quarterfinals':0, 'Semifinals': 0, 'Final':0,\n",
    "                                 'WeightedSum':0})\n",
    "genSerie = pd.Series({'1st Round':0, '2nd Round': 0, '3rd Round':0, '4th Round':0,'Quarterfinals':0,'Round Robin':0,\n",
    "          'Semifinals':0,'The Final':0 })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Player</th>\n",
       "      <th>FirstRounds</th>\n",
       "      <th>Quarterfinals</th>\n",
       "      <th>Semifinals</th>\n",
       "      <th>Final</th>\n",
       "      <th>WeightedSum</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Federer R.</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Player  FirstRounds  Quarterfinals  Semifinals  Final  WeightedSum\n",
       "0   Dosedel S.            0              0           0      0            0\n",
       "1   Enqvist T.            0              0           0      0            0\n",
       "2    Escude N.            0              0           0      0            0\n",
       "3   Federer R.            0              0           0      0            0\n",
       "4  Fromberg R.            0              0           0      0            0"
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_Unique_Players.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [],
   "source": [
    "def roundWin(data = df_atp_Train, genSerie = genSerie, targetData = df_Unique_Players, inplace = True):\n",
    "    \"\"\"\n",
    "    Overview : \n",
    "    Filling the target data with rounds victories rate per player and per round type.\n",
    "    \n",
    "    #Arguments : \n",
    "    data : dataframe of players matches having at least the three columns : 'Winner','Loser','Round'\n",
    "    genSerie : pre-defined Serie with a standard format for local (within function) computations\n",
    "    targetData : dataframe to be filled with per round and player victories rates\n",
    "    inplace : boolean, whether to fill the passed target data or return a filled copy.\n",
    "    \n",
    "    \n",
    "    #Returns:\n",
    "    Returns the filled target data (if inplace = False)\n",
    "    Fills directly the target data passed as argument (if inplace = True)\n",
    "    \n",
    "    \"\"\"\n",
    "    \n",
    "    progress = 0\n",
    "    playersLen = len(targetData)\n",
    "    tenPer, per = playersLen//10, 0\n",
    "    \n",
    "    roundWeights = np.array([0.1,0.2,0.3,0.4,0.6,0,0.7,0.9])  #Weights to be attributed to victories rate per round\n",
    "                                                              #The weights have ascending values (importance of rounds)\n",
    "    \n",
    "    if(not inplace):\n",
    "        targetData = targetData.copy()\n",
    "        \n",
    "    for i, player in enumerate(targetData['Player']):\n",
    "        resultSerie = genSerie.copy()\n",
    "        playerW = data.loc[data['Winner']==player,'Round']\n",
    "        playerL = data.loc[data['Loser']==player,'Round']\n",
    "        \n",
    "        perRoundVictories = playerW.value_counts()\n",
    "        perRoundLosses = playerL.value_counts()\n",
    "        \n",
    "        totalResults = perRoundVictories+perRoundLosses\n",
    "        victoryFractions = perRoundVictories/totalResults\n",
    "        \n",
    "        resultSerie[victoryFractions.keys()] = victoryFractions\n",
    "        resultSerie[resultSerie.isna()] = 0\n",
    "        \n",
    "        FirstRounds = np.mean(resultSerie[0:4].values)    #Mean victories rate of the first 4 rounds\n",
    "        Quarterfinals = resultSerie['Quarterfinals']      #Quarterfinals victories rate\n",
    "        Semifinals = resultSerie['Semifinals']            #Semifinals victories rate\n",
    "        Final = resultSerie['The Final']                  #Final victories rate\n",
    "        \n",
    "        weightedVals = np.sum(roundWeights * resultSerie.values) \n",
    "        \n",
    "        targetData.loc[i,'FirstRounds'] = FirstRounds\n",
    "        targetData.loc[i,'Quarterfinals'] = Quarterfinals\n",
    "        targetData.loc[i,'Semifinals'] = Semifinals\n",
    "        targetData.loc[i,'Final'] = Final\n",
    "        targetData.loc[i,'WeightedSum'] = weightedVals\n",
    "        \n",
    "        \n",
    "        if(per==progress):\n",
    "            print(\"Computations in progress \"+str( np.round(100*per/playersLen, decimals=0))+\"%\")\n",
    "            per += tenPer\n",
    "        progress += 1\n",
    "        if(progress==playersLen):\n",
    "            print(\"Done !\")\n",
    "            \n",
    "    if(not inplace):\n",
    "        return(targetData)\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computations in progress 0.0%\n",
      "Computations in progress 10.0%\n",
      "Computations in progress 20.0%\n",
      "Computations in progress 30.0%\n",
      "Computations in progress 40.0%\n",
      "Computations in progress 50.0%\n",
      "Computations in progress 60.0%\n",
      "Computations in progress 70.0%\n",
      "Computations in progress 80.0%\n",
      "Computations in progress 90.0%\n",
      "Computations in progress 100.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "roundWin(data = df_atp_Train, genSerie = genSerie, targetData = df_Unique_Players, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Player</th>\n",
       "      <th>FirstRounds</th>\n",
       "      <th>Quarterfinals</th>\n",
       "      <th>Semifinals</th>\n",
       "      <th>Final</th>\n",
       "      <th>WeightedSum</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0.263393</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.192857</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0.461347</td>\n",
       "      <td>0.541667</td>\n",
       "      <td>0.461538</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.501725</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0.473575</td>\n",
       "      <td>0.450000</td>\n",
       "      <td>0.555556</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>1.594823</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Federer R.</td>\n",
       "      <td>0.873383</td>\n",
       "      <td>0.818627</td>\n",
       "      <td>0.750000</td>\n",
       "      <td>0.656716</td>\n",
       "      <td>2.479830</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>0.218833</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.433687</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Player  FirstRounds  Quarterfinals  Semifinals     Final  WeightedSum\n",
       "0   Dosedel S.     0.263393       0.000000    0.000000  0.000000     0.192857\n",
       "1   Enqvist T.     0.461347       0.541667    0.461538  0.500000     1.501725\n",
       "2    Escude N.     0.473575       0.450000    0.555556  0.600000     1.594823\n",
       "3   Federer R.     0.873383       0.818627    0.750000  0.656716     2.479830\n",
       "4  Fromberg R.     0.218833       0.500000    0.000000  0.000000     0.433687"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_Unique_Players.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train['WeightedRndDiff'] = 0   #Depends only on the the two players\n",
    "df_atp_Train['RoundDiff'] = 0         #depends on the two players + the round of the match"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "def RoundDiff(sourceData = df_Unique_Players, targetData = df_atp_Train, forTestData = False):\n",
    "    \"\"\"\n",
    "    Overview :\n",
    "    Values Filling of the 'WeightedRndDiff' and 'RoundDiff' features for both train and test set\n",
    "    \n",
    "    Arguments : \n",
    "    sourceData : dataframe of all players with their corresponding per round victories rate\n",
    "    targetData : dataframe having at least the 3 features 'Player1','Player2','Round' to be filled \n",
    "                with players victories rate per round type\n",
    "    forTestData : boolean, whether it's applied to the test or train data (the main difference is that test data\n",
    "                  may contain new players with no recorded per round victories rate, we firt add them to a copy\n",
    "                  of the sourceData with 0 everywhere before filling the target data)\n",
    "    \n",
    "    Returns :\n",
    "    inPlace filling of 'WeightedRndDiff' and 'RoundDiff' features\n",
    "    \n",
    "    \n",
    "    \"\"\"\n",
    "    \n",
    "    #Test Data Case\n",
    "    if(forTestData):\n",
    "        sourceData = sourceData.copy()\n",
    "        uniquePlayersT =targetData.loc[:,['Winner','Loser']].unstack().unique()\n",
    "        df_Unique_Players_T = pd.DataFrame({'Player':uniquePlayersT,'FirstRounds':0, \n",
    "                                  'Quarterfinals':0, 'Semifinals': 0, 'Final':0,\n",
    "                                 'WeightedSum':0})\n",
    "        newPlayersInd = df_Unique_Players_T['Player'].apply(lambda x : x not in df_Unique_Players['Player'].values)\n",
    "        df_Unique_Players_T = df_Unique_Players_T.loc[newPlayersInd,:]\n",
    "        sourceData = pd.concat([df_Unique_Players,df_Unique_Players_T], axis=0)\n",
    "        \n",
    "    \n",
    "    progress = 0\n",
    "    playersLen = len(targetData)\n",
    "    tenPer, per = playersLen//10, 0\n",
    "    \n",
    "    for i in range(len(targetData)):\n",
    "        p1 = targetData.loc[targetData.index[i],'Player1']\n",
    "        p2 = targetData.loc[targetData.index[i],'Player2']\n",
    "        matchRound = targetData.loc[targetData.index[i],'Round']\n",
    "        \n",
    "        wSump1 = sourceData.loc[sourceData['Player']==p1,'WeightedSum']\n",
    "        wSump2 = sourceData.loc[sourceData['Player']==p2,'WeightedSum']\n",
    "                    \n",
    "        \n",
    "        diffW = wSump1.values - wSump2.values\n",
    "        targetData.loc[targetData.index[i],'WeightedRndDiff'] = diffW\n",
    "        \n",
    "        if 'Round' in matchRound:\n",
    "            roundValP1 = sourceData.loc[sourceData['Player']==p1,'FirstRounds']\n",
    "            roundValP2 = sourceData.loc[sourceData['Player']==p2,'FirstRounds']\n",
    "        elif matchRound == 'Quarterfinals':\n",
    "            roundValP1 = sourceData.loc[sourceData['Player']==p1,'Quarterfinals']\n",
    "            roundValP2 = sourceData.loc[sourceData['Player']==p2,'Quarterfinals']\n",
    "        elif matchRound == 'Semifinals':\n",
    "            roundValP1 = sourceData.loc[sourceData['Player']==p1,'Semifinals']\n",
    "            roundValP2 = sourceData.loc[sourceData['Player']==p2,'Semifinals']\n",
    "        elif matchRound == 'The Final':\n",
    "            roundValP1 = sourceData.loc[sourceData['Player']==p1,'Final']\n",
    "            roundValP2 = sourceData.loc[sourceData['Player']==p2,'Final']\n",
    "        else:\n",
    "            roundValP1 = wSump1\n",
    "            roundValP2 = wSump2\n",
    "        \n",
    "        diff = roundValP1.values - roundValP2.values\n",
    "        targetData.loc[targetData.index[i],'RoundDiff'] = diff\n",
    "        \n",
    "            \n",
    "        if(per==progress):\n",
    "            print(\"Computations in progress \"+str( np.round(100*per/playersLen, decimals=0))+\"%\")\n",
    "            per += tenPer\n",
    "        progress += 1\n",
    "        if(progress==playersLen):\n",
    "            print(\"Done !\")\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computations in progress 0.0%\n",
      "Computations in progress 10.0%\n",
      "Computations in progress 20.0%\n",
      "Computations in progress 30.0%\n",
      "Computations in progress 40.0%\n",
      "Computations in progress 50.0%\n",
      "Computations in progress 60.0%\n",
      "Computations in progress 70.0%\n",
      "Computations in progress 80.0%\n",
      "Computations in progress 90.0%\n",
      "Computations in progress 100.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "RoundDiff(sourceData = df_Unique_Players, targetData = df_atp_Train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>...</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "      <th>WeightedRndDiff</th>\n",
       "      <th>RoundDiff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-14</td>\n",
       "      <td>-107</td>\n",
       "      <td>1.471255</td>\n",
       "      <td>0.341504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-0.5</td>\n",
       "      <td>-0.200000</td>\n",
       "      <td>-51</td>\n",
       "      <td>-5040</td>\n",
       "      <td>-0.062583</td>\n",
       "      <td>-0.003199</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-615</td>\n",
       "      <td>-989</td>\n",
       "      <td>-1.566251</td>\n",
       "      <td>-0.402146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>22</td>\n",
       "      <td>148</td>\n",
       "      <td>2.422362</td>\n",
       "      <td>0.752442</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>117</td>\n",
       "      <td>371</td>\n",
       "      <td>0.346782</td>\n",
       "      <td>0.037285</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 23 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament    ...      \\\n",
       "0  International    Hard  Australian Hardcourt Championships    ...       \n",
       "1  International    Hard  Australian Hardcourt Championships    ...       \n",
       "2  International    Hard  Australian Hardcourt Championships    ...       \n",
       "3  International    Hard  Australian Hardcourt Championships    ...       \n",
       "4  International    Hard  Australian Hardcourt Championships    ...       \n",
       "\n",
       "         Player1         Player2 Won  Year p1Vsp2   Wp1Vsp2  RankDiff  \\\n",
       "0    Ljubicic I.      Dosedel S.   0  2000   -1.0 -0.333333       -14   \n",
       "1     Clement A.      Enqvist T.   0  2000   -0.5 -0.200000       -51   \n",
       "2  Baccanello P.       Escude N.   0  2000   -1.0 -0.333333      -615   \n",
       "3     Federer R.  Knippschild J.   1  2000    1.0  0.333333        22   \n",
       "4    Fromberg R.   Woodbridge T.   1  2000    1.0  0.333333       117   \n",
       "\n",
       "   PtsDiff  WeightedRndDiff  RoundDiff  \n",
       "0     -107         1.471255   0.341504  \n",
       "1    -5040        -0.062583  -0.003199  \n",
       "2     -989        -1.566251  -0.402146  \n",
       "3      148         2.422362   0.752442  \n",
       "4      371         0.346782   0.037285  \n",
       "\n",
       "[5 rows x 23 columns]"
      ]
     },
     "execution_count": 159,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's visualize some scatter plots labeled with the output value (0-1) to see the effect of the added features related to performance difference :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (282023379)>"
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pln.ggplot(data=df_atp_Train,mapping= pln.aes(x='RankDiff',y='PtsDiff', colour='factor(Won)'))+pln.geom_point(alpha=0.7)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We do the same for test data (based on computed values of the train examples of course)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Test['WeightedRndDiff'] = 0   #Depends only on the the two players\n",
    "df_atp_Test['RoundDiff'] = 0         #depends on the two players + the round of the match"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computations in progress 0.0%\n",
      "Computations in progress 10.0%\n",
      "Computations in progress 20.0%\n",
      "Computations in progress 30.0%\n",
      "Computations in progress 40.0%\n",
      "Computations in progress 50.0%\n",
      "Computations in progress 60.0%\n",
      "Computations in progress 70.0%\n",
      "Computations in progress 80.0%\n",
      "Computations in progress 90.0%\n",
      "Computations in progress 100.0%\n",
      "Done !\n"
     ]
    }
   ],
   "source": [
    "RoundDiff(sourceData = df_Unique_Players, targetData = df_atp_Test, forTestData=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Best of</th>\n",
       "      <th>Court</th>\n",
       "      <th>Date</th>\n",
       "      <th>LPts</th>\n",
       "      <th>LRank</th>\n",
       "      <th>Loser</th>\n",
       "      <th>Round</th>\n",
       "      <th>Series</th>\n",
       "      <th>Surface</th>\n",
       "      <th>Tournament</th>\n",
       "      <th>...</th>\n",
       "      <th>Player1</th>\n",
       "      <th>Player2</th>\n",
       "      <th>Won</th>\n",
       "      <th>Year</th>\n",
       "      <th>p1Vsp2</th>\n",
       "      <th>Wp1Vsp2</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "      <th>WeightedRndDiff</th>\n",
       "      <th>RoundDiff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>648</td>\n",
       "      <td>77</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Ljubicic I.</td>\n",
       "      <td>Dosedel S.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-14</td>\n",
       "      <td>-107</td>\n",
       "      <td>1.471255</td>\n",
       "      <td>0.341504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>818</td>\n",
       "      <td>56</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Clement A.</td>\n",
       "      <td>Enqvist T.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-0.5</td>\n",
       "      <td>-0.200000</td>\n",
       "      <td>-51</td>\n",
       "      <td>-5040</td>\n",
       "      <td>-0.062583</td>\n",
       "      <td>-0.003199</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>36</td>\n",
       "      <td>655</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Baccanello P.</td>\n",
       "      <td>Escude N.</td>\n",
       "      <td>0</td>\n",
       "      <td>2000</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-0.333333</td>\n",
       "      <td>-615</td>\n",
       "      <td>-989</td>\n",
       "      <td>-1.566251</td>\n",
       "      <td>-0.402146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>588</td>\n",
       "      <td>87</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Federer R.</td>\n",
       "      <td>Knippschild J.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>22</td>\n",
       "      <td>148</td>\n",
       "      <td>2.422362</td>\n",
       "      <td>0.752442</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>2000-01-03</td>\n",
       "      <td>254</td>\n",
       "      <td>198</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>International</td>\n",
       "      <td>Hard</td>\n",
       "      <td>Australian Hardcourt Championships</td>\n",
       "      <td>...</td>\n",
       "      <td>Fromberg R.</td>\n",
       "      <td>Woodbridge T.</td>\n",
       "      <td>1</td>\n",
       "      <td>2000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>117</td>\n",
       "      <td>371</td>\n",
       "      <td>0.346782</td>\n",
       "      <td>0.037285</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 23 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Best of    Court       Date  LPts  LRank           Loser      Round  \\\n",
       "0        3  Outdoor 2000-01-03   648     77     Ljubicic I.  1st Round   \n",
       "1        3  Outdoor 2000-01-03   818     56      Clement A.  1st Round   \n",
       "2        3  Outdoor 2000-01-03    36    655   Baccanello P.  1st Round   \n",
       "3        3  Outdoor 2000-01-03   588     87  Knippschild J.  1st Round   \n",
       "4        3  Outdoor 2000-01-03   254    198   Woodbridge T.  1st Round   \n",
       "\n",
       "          Series Surface                          Tournament    ...      \\\n",
       "0  International    Hard  Australian Hardcourt Championships    ...       \n",
       "1  International    Hard  Australian Hardcourt Championships    ...       \n",
       "2  International    Hard  Australian Hardcourt Championships    ...       \n",
       "3  International    Hard  Australian Hardcourt Championships    ...       \n",
       "4  International    Hard  Australian Hardcourt Championships    ...       \n",
       "\n",
       "         Player1         Player2 Won  Year p1Vsp2   Wp1Vsp2  RankDiff  \\\n",
       "0    Ljubicic I.      Dosedel S.   0  2000   -1.0 -0.333333       -14   \n",
       "1     Clement A.      Enqvist T.   0  2000   -0.5 -0.200000       -51   \n",
       "2  Baccanello P.       Escude N.   0  2000   -1.0 -0.333333      -615   \n",
       "3     Federer R.  Knippschild J.   1  2000    1.0  0.333333        22   \n",
       "4    Fromberg R.   Woodbridge T.   1  2000    1.0  0.333333       117   \n",
       "\n",
       "   PtsDiff  WeightedRndDiff  RoundDiff  \n",
       "0     -107         1.471255   0.341504  \n",
       "1    -5040        -0.062583  -0.003199  \n",
       "2     -989        -1.566251  -0.402146  \n",
       "3      148         2.422362   0.752442  \n",
       "4      371         0.346782   0.037285  \n",
       "\n",
       "[5 rows x 23 columns]"
      ]
     },
     "execution_count": 163,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Best of', 'Court', 'Date', 'LPts', 'LRank', 'Loser', 'Round', 'Series',\n",
       "       'Surface', 'Tournament', 'WPts', 'WRank', 'Winner', 'Player1',\n",
       "       'Player2', 'Won', 'Year', 'p1Vsp2', 'Wp1Vsp2', 'RankDiff', 'PtsDiff',\n",
       "       'WeightedRndDiff', 'RoundDiff'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 164,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_atp_Train.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [],
   "source": [
    "featuresCols = np.array(['Won','Year','Court','Round','Surface','RankDiff','PtsDiff','WeightedRndDiff','RoundDiff' ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [],
   "source": [
    "TrainData = df_atp_Train.loc[:,featuresCols].copy()\n",
    "TrainX = TrainData.loc[:,'Won'!=featuresCols]\n",
    "TrainY = TrainData.loc[:,'Won']\n",
    "\n",
    "TestData = df_atp_Test.loc[:,featuresCols].copy()\n",
    "TestX = TestData.loc[:,'Won'!=featuresCols]\n",
    "TestY = TestData.loc[:,'Won']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_atp_Train.to_csv(date_format='%Y-%m-%d',header=True, path_or_buf='ATPTrain.csv')\n",
    "df_atp_Test.to_csv(date_format='%Y-%m-%d',header=True, path_or_buf='ATPTest.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Categorical variables to dummy variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import preprocessing\n",
    "from sklearn.metrics import classification_report, confusion_matrix  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Year', 'Court', 'Round', 'Surface', 'RankDiff', 'PtsDiff',\n",
       "       'WeightedRndDiff', 'RoundDiff'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 169,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TrainX.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CourtCat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Outdoor</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Indoor</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  CourtCat\n",
       "0  Outdoor\n",
       "1   Indoor"
      ]
     },
     "execution_count": 170,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "courtCats = TrainX.Court.unique()\n",
    "CourtCat = pd.DataFrame({'CourtCat':TrainX.Court.unique()})\n",
    "CourtCat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>SurfaceCat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Hard</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Clay</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Grass</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Carpet</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  SurfaceCat\n",
       "0       Hard\n",
       "1       Clay\n",
       "2      Grass\n",
       "3     Carpet"
      ]
     },
     "execution_count": 171,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "surfCats = TrainX.Surface.unique()\n",
    "SurfaceCategories = pd.DataFrame({'SurfaceCat':TrainX.Surface.unique()})\n",
    "SurfaceCategories"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>RoundCat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1st Round</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2nd Round</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Quarterfinals</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Semifinals</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>The Final</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>3rd Round</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>4th Round</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Round Robin</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        RoundCat\n",
       "0      1st Round\n",
       "1      2nd Round\n",
       "2  Quarterfinals\n",
       "3     Semifinals\n",
       "4      The Final\n",
       "5      3rd Round\n",
       "6      4th Round\n",
       "7    Round Robin"
      ]
     },
     "execution_count": 172,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roundCats = TrainX.Round.unique()\n",
    "RoundCategories = pd.DataFrame({'RoundCat':TrainX.Round.unique()})\n",
    "RoundCategories"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We turn these features into 0-1 encoding (Dummy variables)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [],
   "source": [
    "surfaceDummiesTrain = pd.get_dummies(TrainX,columns= np.array(['Surface']), prefix='', prefix_sep='')\n",
    "surfaceDummiesTrain = surfaceDummiesTrain.T.reindex(surfCats).T.fillna(0)\n",
    "surfaceDummiesTrain = surfaceDummiesTrain.loc[:,surfCats]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [],
   "source": [
    "surfaceDummiesTest = pd.get_dummies(TestX,columns= np.array(['Surface']), prefix='', prefix_sep='')\n",
    "surfaceDummiesTest = surfaceDummiesTest.T.reindex(surfCats).T.fillna(0)\n",
    "surfaceDummiesTest = surfaceDummiesTest.loc[:,surfCats]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {},
   "outputs": [],
   "source": [
    "courtDummiesTrain = pd.get_dummies(TrainX,columns= np.array(['Court']), prefix='', prefix_sep='')\n",
    "courtDummiesTrain = courtDummiesTrain.T.reindex(courtCats).T.fillna(0)\n",
    "courtDummiesTrain = courtDummiesTrain.loc[:,courtCats]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {},
   "outputs": [],
   "source": [
    "courtDummiesTest = pd.get_dummies(TestX,columns= np.array(['Court']), prefix='', prefix_sep='')\n",
    "courtDummiesTest = courtDummiesTest.T.reindex(courtCats).T.fillna(0)\n",
    "courtDummiesTest = courtDummiesTest.loc[:,courtCats]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {},
   "outputs": [],
   "source": [
    "roundDummiesTrain = pd.get_dummies(TrainX,columns= np.array(['Round']), prefix='', prefix_sep='')\n",
    "roundDummiesTrain = roundDummiesTrain.T.reindex(roundCats).T.fillna(0)\n",
    "roundDummiesTrain = roundDummiesTrain.loc[:,roundCats]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [],
   "source": [
    "roundDummiesTest = pd.get_dummies(TestX,columns= np.array(['Round']), prefix='', prefix_sep='')\n",
    "roundDummiesTest = roundDummiesTest.T.reindex(roundCats).T.fillna(0)\n",
    "roundDummiesTest = roundDummiesTest.loc[:,roundCats]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Concatenate the whole\n",
    "TrainX = pd.concat([TrainX, surfaceDummiesTrain, courtDummiesTrain, roundDummiesTrain], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Year</th>\n",
       "      <th>Court</th>\n",
       "      <th>Round</th>\n",
       "      <th>Surface</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "      <th>WeightedRndDiff</th>\n",
       "      <th>RoundDiff</th>\n",
       "      <th>Hard</th>\n",
       "      <th>Clay</th>\n",
       "      <th>...</th>\n",
       "      <th>Outdoor</th>\n",
       "      <th>Indoor</th>\n",
       "      <th>1st Round</th>\n",
       "      <th>2nd Round</th>\n",
       "      <th>Quarterfinals</th>\n",
       "      <th>Semifinals</th>\n",
       "      <th>The Final</th>\n",
       "      <th>3rd Round</th>\n",
       "      <th>4th Round</th>\n",
       "      <th>Round Robin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2000</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-14</td>\n",
       "      <td>-107</td>\n",
       "      <td>1.471255</td>\n",
       "      <td>0.341504</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2000</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-51</td>\n",
       "      <td>-5040</td>\n",
       "      <td>-0.062583</td>\n",
       "      <td>-0.003199</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2000</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-615</td>\n",
       "      <td>-989</td>\n",
       "      <td>-1.566251</td>\n",
       "      <td>-0.402146</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2000</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>22</td>\n",
       "      <td>148</td>\n",
       "      <td>2.422362</td>\n",
       "      <td>0.752442</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2000</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>117</td>\n",
       "      <td>371</td>\n",
       "      <td>0.346782</td>\n",
       "      <td>0.037285</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Year    Court      Round Surface  RankDiff  PtsDiff  WeightedRndDiff  \\\n",
       "0  2000  Outdoor  1st Round    Hard       -14     -107         1.471255   \n",
       "1  2000  Outdoor  1st Round    Hard       -51    -5040        -0.062583   \n",
       "2  2000  Outdoor  1st Round    Hard      -615     -989        -1.566251   \n",
       "3  2000  Outdoor  1st Round    Hard        22      148         2.422362   \n",
       "4  2000  Outdoor  1st Round    Hard       117      371         0.346782   \n",
       "\n",
       "   RoundDiff  Hard  Clay     ...       Outdoor  Indoor  1st Round  2nd Round  \\\n",
       "0   0.341504     1     0     ...             1       0          1          0   \n",
       "1  -0.003199     1     0     ...             1       0          1          0   \n",
       "2  -0.402146     1     0     ...             1       0          1          0   \n",
       "3   0.752442     1     0     ...             1       0          1          0   \n",
       "4   0.037285     1     0     ...             1       0          1          0   \n",
       "\n",
       "   Quarterfinals  Semifinals  The Final  3rd Round  4th Round  Round Robin  \n",
       "0              0           0          0          0          0            0  \n",
       "1              0           0          0          0          0            0  \n",
       "2              0           0          0          0          0            0  \n",
       "3              0           0          0          0          0            0  \n",
       "4              0           0          0          0          0            0  \n",
       "\n",
       "[5 rows x 22 columns]"
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TrainX.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "metadata": {},
   "outputs": [],
   "source": [
    "TestX = pd.concat([TestX, surfaceDummiesTest, courtDummiesTest, roundDummiesTest], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Year</th>\n",
       "      <th>Court</th>\n",
       "      <th>Round</th>\n",
       "      <th>Surface</th>\n",
       "      <th>RankDiff</th>\n",
       "      <th>PtsDiff</th>\n",
       "      <th>WeightedRndDiff</th>\n",
       "      <th>RoundDiff</th>\n",
       "      <th>Hard</th>\n",
       "      <th>Clay</th>\n",
       "      <th>...</th>\n",
       "      <th>Outdoor</th>\n",
       "      <th>Indoor</th>\n",
       "      <th>1st Round</th>\n",
       "      <th>2nd Round</th>\n",
       "      <th>Quarterfinals</th>\n",
       "      <th>Semifinals</th>\n",
       "      <th>The Final</th>\n",
       "      <th>3rd Round</th>\n",
       "      <th>4th Round</th>\n",
       "      <th>Round Robin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47028</th>\n",
       "      <td>2017</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-81</td>\n",
       "      <td>-317</td>\n",
       "      <td>0.094949</td>\n",
       "      <td>0.112374</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47029</th>\n",
       "      <td>2017</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-15</td>\n",
       "      <td>-248</td>\n",
       "      <td>-1.052684</td>\n",
       "      <td>-0.021857</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47030</th>\n",
       "      <td>2017</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>5</td>\n",
       "      <td>320</td>\n",
       "      <td>0.376746</td>\n",
       "      <td>0.248541</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47031</th>\n",
       "      <td>2017</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-96</td>\n",
       "      <td>-558</td>\n",
       "      <td>-0.216590</td>\n",
       "      <td>-0.077189</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47032</th>\n",
       "      <td>2017</td>\n",
       "      <td>Outdoor</td>\n",
       "      <td>1st Round</td>\n",
       "      <td>Hard</td>\n",
       "      <td>-16</td>\n",
       "      <td>-715</td>\n",
       "      <td>-0.293613</td>\n",
       "      <td>-0.136051</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       Year    Court      Round Surface  RankDiff  PtsDiff  WeightedRndDiff  \\\n",
       "47028  2017  Outdoor  1st Round    Hard       -81     -317         0.094949   \n",
       "47029  2017  Outdoor  1st Round    Hard       -15     -248        -1.052684   \n",
       "47030  2017  Outdoor  1st Round    Hard         5      320         0.376746   \n",
       "47031  2017  Outdoor  1st Round    Hard       -96     -558        -0.216590   \n",
       "47032  2017  Outdoor  1st Round    Hard       -16     -715        -0.293613   \n",
       "\n",
       "       RoundDiff  Hard  Clay     ...       Outdoor  Indoor  1st Round  \\\n",
       "47028   0.112374     1     0     ...             1       0          1   \n",
       "47029  -0.021857     1     0     ...             1       0          1   \n",
       "47030   0.248541     1     0     ...             1       0          1   \n",
       "47031  -0.077189     1     0     ...             1       0          1   \n",
       "47032  -0.136051     1     0     ...             1       0          1   \n",
       "\n",
       "       2nd Round  Quarterfinals  Semifinals  The Final  3rd Round  4th Round  \\\n",
       "47028          0              0           0          0          0          0   \n",
       "47029          0              0           0          0          0          0   \n",
       "47030          0              0           0          0          0          0   \n",
       "47031          0              0           0          0          0          0   \n",
       "47032          0              0           0          0          0          0   \n",
       "\n",
       "       Round Robin  \n",
       "47028            0  \n",
       "47029            0  \n",
       "47030            0  \n",
       "47031            0  \n",
       "47032            0  \n",
       "\n",
       "[5 rows x 22 columns]"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TestX.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [],
   "source": [
    "TrainX.drop(axis=1,columns=['Court','Surface', 'Round'], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "metadata": {},
   "outputs": [],
   "source": [
    "TestX.drop(axis=1,columns=['Court','Surface','Round'], inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Modeling  \n",
    "We will use a random forest implementation in sklearn to fit our model, with some hyper-parameters tuning for a better estimation of the optimal model.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Import Random Forest Model\n",
    "from sklearn.ensemble import RandomForestClassifier , AdaBoostClassifier, GradientBoostingClassifier\n",
    "from sklearn.model_selection import RandomizedSearchCV, GridSearchCV\n",
    "from sklearn.tree import DecisionTreeClassifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "metadata": {},
   "outputs": [],
   "source": [
    "rf_Fit0=RandomForestClassifier(n_estimators=1000, random_state=1995, min_samples_leaf=20, max_features=9, criterion='gini' )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features=9, max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=20, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=1000, n_jobs=1,\n",
       "            oob_score=False, random_state=1995, verbose=0,\n",
       "            warm_start=False)"
      ]
     },
     "execution_count": 187,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_Fit0.fit(TrainX,TrainY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "metadata": {},
   "outputs": [],
   "source": [
    "rf_pred = rf_Fit0.predict(TestX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6444022770398482"
      ]
     },
     "execution_count": 199,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_Fit0.score(TestX, TestY)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tuning Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Number of trees in random forest\n",
    "n_estimators = np.linspace(start = 1200, stop = 3000, num = 10, dtype='int32')\n",
    "#Number of features to consider at every split\n",
    "max_features = [5,6,7,8,10]\n",
    "#Maximum number of levels in tree\n",
    "max_depth = list(np.linspace(30, 80, num = 5, dtype='int32'))\n",
    "max_depth.append(None)\n",
    "#Minimum number of samples required to split a node\n",
    "min_samples_split = [10, 50, 80]\n",
    "#Minimum number of samples required at each leaf node\n",
    "min_samples_leaf = [5, 10, 40]\n",
    "#Method of selecting samples for training each tree\n",
    "bootstrap = [True, False]\n",
    "#Information critereon  \n",
    "criterion=['gini']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Random grid\n",
    "random_grid = {'n_estimators': n_estimators,\n",
    "               'max_features': max_features,\n",
    "               'max_depth': max_depth,\n",
    "               'min_samples_split': min_samples_split,\n",
    "               'min_samples_leaf': min_samples_leaf,\n",
    "               'bootstrap': bootstrap,\n",
    "               'criterion' : criterion\n",
    "              }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Random forest Base model to tune\n",
    "rf_fit = RandomForestClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {},
   "outputs": [],
   "source": [
    "rf_random_fit = RandomizedSearchCV(estimator = rf_fit, param_distributions = random_grid, n_iter = 30, cv = 3, random_state=1995, n_jobs = -1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [],
   "source": [
    "TrainY = TrainY.squeeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=3, error_score='raise',\n",
       "          estimator=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,\n",
       "            oob_score=False, random_state=None, verbose=0,\n",
       "            warm_start=False),\n",
       "          fit_params=None, iid=True, n_iter=30, n_jobs=-1,\n",
       "          param_distributions={'n_estimators': array([1200, 1400, 1600, 1800, 2000, 2200, 2400, 2600, 2800, 3000],\n",
       "      dtype=int32), 'max_features': [5, 6, 7, 8, 10], 'max_depth': [30, 42, 55, 67, 80, None], 'min_samples_split': [10, 50, 80], 'min_samples_leaf': [5, 10, 40], 'bootstrap': [True, False], 'criterion': ['gini']},\n",
       "          pre_dispatch='2*n_jobs', random_state=1995, refit=True,\n",
       "          return_train_score='warn', scoring=None, verbose=0)"
      ]
     },
     "execution_count": 195,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fit the random search model\n",
    "rf_random_fit.fit(TrainX, TrainY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {},
   "outputs": [],
   "source": [
    "best_rf = rf_random_fit.best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=67, max_features=5, max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=40, min_samples_split=80,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=2600, n_jobs=1,\n",
       "            oob_score=False, random_state=None, verbose=0,\n",
       "            warm_start=False)"
      ]
     },
     "execution_count": 197,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_rf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6432637571157496"
      ]
     },
     "execution_count": 198,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_rf.score(X=TestX,y=TestY)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We get a test accuracy of 64.4%. Let's have a look at the variable importance according to this model.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_imp = pd.Series(best_rf.feature_importances_, index=TrainX.columns).sort_values(ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Features importance in the optimal Random Forest estimator')"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10,6))\n",
    "sns.barplot(x=feature_imp, y=feature_imp.index)\n",
    "plt.title('Features importance in the optimal Random Forest estimator')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [],
   "source": [
    "bestPredictions = best_rf.predict(TestX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {},
   "outputs": [],
   "source": [
    "bestPredictions = bestPredictions.reshape((len(bestPredictions),1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [],
   "source": [
    "trueValues = np.array(TestY).reshape((len(TestY),1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[872, 482],\n",
       "       [458, 823]])"
      ]
     },
     "execution_count": 205,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Confusion matrix\n",
    "confusion_matrix(y_pred=bestPredictions, y_true=trueValues)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The accuracy of 64% we've got is somewhat mediocre and can be explained by the fact that the nature of the problem is not an ordinary ML task since it includes the Time factor that we didn't explore using Time Series technics. The lack of further features is also a crucial factor, and eventhough we were able to use betting data and that would increase significantly with no doubt our prediction accuracy, the idea itself lacks honesty because it's the output of other peoples' models, and thus it doesn't make sense to use it and pretend having built a realistic ML model. "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}