Proposed solution for ratings calc
13 Apr '05 21:52I think something needs to be done about large rating variations.
Some deliberately lower their rating by resigning games or allowing timeouts so that they can enter lower banded tournies. Others, maybe though no fault of their own are absent for sometime and thus suffer move timeout losses. This is the more likely senario.
Playing someone rated 1100 who has been as high as 1700 in a banded tourni is a bit gauling as it reduces your rating far more than it should do given the players skill level. A floor would seem to be the solution but an algorithm of rolling average/highest rating would seem the most appropriate. It may be difficult to set up but would be fairer.
(Highest rating - Rolling Average rating) /2 = floor differential.
Floor = Rolling average - floor differential
This would mean that a player rated consistantly in the 2000's over 100 games may have a rating of 2100 and a rolling average of 2010 thus his floor would be 1975.
A player who's rating goes up and down due to grouped wins/loses or because they have progressed due to good fortune may have a max rating of 1600 and a rolling average of 1350 thus their floor would be 1225. This would be for someone normally rated in the 1300's but had a spike up to 1600. Thus 1225 would be only 100 lower than their normal rating.
I know it's a bit complicated but it would result in players who are missing from the site returning at a rating level consistant with their skill level.
🙄
Some deliberately lower their rating by resigning games or allowing timeouts so that they can enter lower banded tournies. Others, maybe though no fault of their own are absent for sometime and thus suffer move timeout losses. This is the more likely senario.
Playing someone rated 1100 who has been as high as 1700 in a banded tourni is a bit gauling as it reduces your rating far more than it should do given the players skill level. A floor would seem to be the solution but an algorithm of rolling average/highest rating would seem the most appropriate. It may be difficult to set up but would be fairer.
(Highest rating - Rolling Average rating) /2 = floor differential.
Floor = Rolling average - floor differential
This would mean that a player rated consistantly in the 2000's over 100 games may have a rating of 2100 and a rolling average of 2010 thus his floor would be 1975.
A player who's rating goes up and down due to grouped wins/loses or because they have progressed due to good fortune may have a max rating of 1600 and a rolling average of 1350 thus their floor would be 1225. This would be for someone normally rated in the 1300's but had a spike up to 1600. Thus 1225 would be only 100 lower than their normal rating.
I know it's a bit complicated but it would result in players who are missing from the site returning at a rating level consistant with their skill level.
🙄
Back to Top