Currently, the Perf(ormance) calculation of a tournament may be quite off, because games where the players are very far apart in their ratings are counted too high.
E.g. with the current method, a player who really has 1200 but plays in a tournament with only 2400's and loses every single game will get a performance of 2000. Which is obviously nonsense.
Maybe the easiest would be to only count games where players are not so far apart. Or count those games only, if the result has got some information. E.g. when the 1200 wins against a 2400, that would be counted.
But then, if the 1200 has won one game against a 2400 and lost 8 against those and I do count only the one game where they won, my calculation would be very far of also.
So I'll implement it using a Glicko-like rating calculation.
Currently, the Perf(ormance) calculation of a tournament may be quite off, because games where the players are very far apart in their ratings are counted too high.
E.g. with the current method, a player who really has 1200 but plays in a tournament with only 2400's and loses every single game will get a performance of 2000. Which is obviously nonsense.
Maybe the easiest would be to only count games where players are not so far apart. Or count those games only, if the result has got some information. E.g. when the 1200 wins against a 2400, that would be counted.
But then, if the 1200 has won one game against a 2400 and lost 8 against those and I do count only the one game where they won, my calculation would be very far of also.
So I'll implement it using a Glicko-like rating calculation.