Enemy AI changes: Change weighting of priorities and add special check #2378
Conversation
|
This should go after #2380 |
fdelapena
added
Has PR Dependencies
The weighting of the priorities has been changed: The action with the highest priority always gets a weight of 10 and actions with a priority of (highest priority - 9) gets a weight of 1. The scaling between is linear. This means if we have two actions, one with a priority of 50 and one with a priority of 45, the first action gets a 10/15 chance and the second action a 5/15 chance to get selected. Moreover a special check has been added if the enemy uses a skill which only heals states without affecting anything else: He actually checks if there is an ally which is affected by at least one state which get healed by the skill. If this is not the case, then the action gets discarded.
|
This PR is still intact - a rebase is not needed after merging #2380. The |
| @@ -244,3 +247,65 @@ const lcf::rpg::EnemyAction* Game_Enemy::ChooseRandomAction() { | |||
| return nullptr; | |||
| } | |||
|
|
|||
| bool Game_Enemy::IsSkillUsableForAI(int skill_id) const { | |||
There was a problem hiding this comment.
RPG_RT does this by calling a function called RPG_Battler::IsSkillEffectiveOn which does a set of checks on each battler. The logic you have here is close but missing some edge cases. Also I need to do more testing to be sure.
|
You've made some good improvements and found some bugs here, but to be honest this is another one like autobattle where it can never be perfect with black box testing. We really need to use RE tools to discover all the edge cases. EnemyAI is next on my list to do as I also want to make a standalone factory function for it with customizable algos, just like autobattle. |
|
Okay, for now I wait for further directions from you regarding this PR to avoid merge conflicts. 👍 |
fdelapena
removed
the
Awaiting Rebase
|
@fmatthew5876 Would it make sense to close this PR in favor of #2381? Because you have already added the priority weight fixes in your PR and you are planning to implement the remaining AI fixes. |
|
Yes I think so, unfortunately |
ghost
deleted the
This PR tries to fix #1831.
This one changes the weighting of the priorities: The action with the highest priority always gets a weight of 10 and actions with a priority of (highest priority - 9) gets a weight of 1. The scaling between is linear. This means if we have two actions, one with a
priority of 50 and one with a priority of 45, the first action gets a 10/15 chance and the second action a 5/15 chance to get selected. Moreover a special check has been added if the enemy uses a skill which only heals states without affecting anything else: He actually checks if there is an ally which is affected by at least one state which get healed by the skill. If this is not the case, then the action gets discarded.