Thursday, October 21, 2004

Ratings magic

magic wand
In a recent post I said:
"It might be reasonable to assume that the recently published 10/04 USCF ratings are a better measure of the players' strength upon entering the Championship than the "official" 8/04 numbers. Of course, it might also be reasonable to believe that ratings mean 'not a nit' in any particular game."
Here is a perspective from someone who has given the topic a lot of thought:

Chess ratings are not devoid of meaning, but they are unquestionably imprecise. As I said last time, we need to keep in mind that FIDE ratings are not the precise measurements we would like them to be. In fact, nobody really knows just how effective the FIDE ratings are at measuring the "true strength" of a player. However, based upon the research that I performed a few years ago in developing my Chessmetrics historical ratings, I don't think it's too far off the mark to say that the estimation error in an Elo rating is normally distributed, with a standard deviation of 50 rating points. That means if somebody has a 2550 rating, we can only be about 70% sure that their "true strength" or "current form" is somewhere between 2500 and 2600. And we can only be 99% sure that their "true strength" is between 2400 and 2700.

I'm sorry about this; I really am. I wish I could tell you that ratings are nice and accurate. I wish I could wave a magic wand and tell you that I have a super-magical rating formula in my bag of tricks, but I don't. In all honesty, I simply don't know who the strongest player in the tournament is. They haven't played enough recent games to let me figure that out with any certainty.

Libya World Championship Statistics (after round 1)" by Jeff Sonas, ChessBase News (web), 6/22/04


Anonymous Anonymous said...

This comment has been removed by a blog administrator.

4:27 PM  

Post a Comment

<< Home