Analyses of the ratings - Spotting the issues

i have an idea to fix this. (i posted it in another thread a few hours ago)

  1. calculate the rating of the team :
    not use simple average anymore, but use an adjusted version of it taking account of unbalanced teams.
    team rating = average rating + 0.5 * standard deviation
    that way, unbalanced teams should be punished, because they get higher team rating.
    (0.5 can be replaced by any number between 0 and 1, depending on how much you want to punish unbalanced teams … but i would recommend something between 0.2 and 0.5 )
    examples :
    team with 1000 and 2000 rated players will get 1750 team rating with 0.5 (1600 with 0.2) (because standard deviation is 500 for them)
    team with 1500 and 1500 will get 1500 team rating (because standard deviation is 0)
  2. calculate elo variation :
    in a same team, everyone should win or lose the same amount (with maybe an exception for players with low number of games) , and the opponent team should lose/win what the first team win/lose (that will prevent the inflation). This should be calculated with the rating of the team, the same way 1v1 rating is made of.
  3. (optionnal) make a one time change of all team ratings to make the average team rating the same of average 1v1 rating by diminishing every rating with the needed amount.
2 Likes