If you have a better predictor than Elo ratings about the probability you will win or lose a chess game, then you can deliberately choose opponents or even choose tournaments to achieve an expected increase in Elo rating: play against the overrated. Given the tremendous attention paid to ratings of professionals nowadays, affecting who will get invited to tournaments and ultimately how much they get paid, there is tremendous incentive to manipulate ratings. A contest elsewhere strongly suggests that there exist better predictors of outcome than current Elo rating.
Is there a way to design rating systems to make this kind of manipulation impossible?
No comments :
Post a Comment