Reader ranking potential issue

[ComeDenechaud]

The automatic implementation of reader ranking (i.e. not manually filling the file) uses number of otoliths read and year of experience to weight the two scores (event specific stock and species as whole). However, there are two potential issues with the scores:

• for institutes with many readers the number read per year is smaller, while in others even a relatively fresh reader taking on the sole responsibility for a stock will get rank one due to the large number of otolith regardless of years of experience
• this stems partly from the current formula for getting the reader rank for event stock, where all three columns are considered together at a similar weight of 1. But because total number of otoliths is simply the product of the previous too, any outlier value will be weighing in twice in the score.
• this is in theory accounted for with the second set of scores based on total reading experience, but since these are only weighed at 0.25 (against 1 for the previous), they don't matter as much

[ComeDenechaud]

[ComeDenechaud]

Example above:

• Reader 5 has only read for 8 years but between 3 and 6 times as many otoliths as the others. It automatically gets rank 2 in this stock despite two other readers having read it for more than 30 years.
• The global score based on total experience adequately gives reader 5 a very low score as it is the least experienced one, but because of unequal weightings it doesn't change much the final assessment and this reader ends up in 2nd overall.

[ComeDenechaud]

Possible solutions:

• change the calculations of the event score: column 3 being a product of columns 1 and 2 gives too much weight to outliers against steady experience, it should either be decoupled from it or alternatively have the event score being only based on total number read and years of experience (the redundancy between a high number per year and the resulting total number read gives too much weight)
• give a higher weight the global score: to offset the influence of newer readers with high numbers per year, give a higher weight than 0.25 to the second score based on total experience

[kbekaert]

Remove third column from the weighing of the score.

[kbekaert]

This was done by Côme during the physical meeting. The excel file must now be made available to the users.