Pages

Sunday, February 6, 2011

Demythologizing the chess player's Elo rating

Chess blog for latest news and trivia (c) Alexandra Kosteniuk, 2011

Hello everyone,


Several different research ideas have been forthcoming about the accuracy and inaccuracy of the Elo rating system. Here is another very interesting and detailed article on www.chessbase.com by Elmer Dumlao Sangalang. He says:



Elmer Dumlao Sangalang

... the Elo Rating of an individual player is not a mathematically precise figure. It is statistically derived with an accuracy in direct correlation to the amount of data (game results) on which it is based. A measurement based on the standard 30 games provides a rating that is 95% probable to be within plus or minus 100 Elo points of its true value. That’s the reason why it is not a certainty that the higher-rated player will always beat his lower-rated opponent. Take this simplistic illustration: Player A rated Elo 2600 plays with Player B rated Elo 2500. The true strength of A lies in the range of 2500 - 2700 while that of B is in the range of 2400 - 2600. If A plays languidly at 2550 while B plays inspired chess at 2550, we will have an even match and the game should result in a draw!


The invention and employment of the Elo rating system may be the best, but not the perfect, thing that ever happened to chess playing and organizing. The late Professor Arpad E. Elo expressed strong sentiments over the inordinate importance being attributed to the rating. He regretted that the Elo Rating System contributed greatly to the prevailing opinion that regards chess as first and foremost a sport. As a result, the chessplayer’s Elo Rating has been overvalued in significance by top-rank players and organizers of major and prestigious chess competitions. And for that matter, even by FIDE, itself.


Go on, read the full article.


From Alexandra Kosteniuk's
www.chessblog.com
Also see her personal blog at
www.chessqueen.com

3 comments:

  1. Hello everyone,

    Whereas the statistical truth is worth speaking, I'd think the important message is to emphasize some fundamental mistakes done when overemphasizing Elo ratings.

    The first is reification (http://en.wikipedia.org/wiki/Reification_(fallacy)), assuming a single number exists which can express the diversity of mental activities involved in playing chess. I'd advise anyone to read Gould's book 'The mis-measure of man' which is a fine debunking of a similar number: the IQ.

    A second, closely related mistake, is to believe that a player's strength exists independently of external factors. Things are changing and people are changing. They might have learned important new (chess) skills recently or are loosing strength from aging (in one or more of the multiple factors mentioned in the first point). One day you play like a 2800 (well if you are say 2600 of course) and on another day, the same person might make moves as week as a 2500 player.

    That is why the rating precision is in fact limited by these natural phenomena and essentially reflects their existence. They'd be a lot more to say on this interesting subject, of course, but that would get too far.

    Cheers,

    Ralph

    ReplyDelete
  2. Saira Fernando, MadridFebruary 7, 2011 at 4:06 AM

    Very well said @Ralph. This is the point I have often made that some players are rated higher only because of the competition they have faced. In any case chess is such a dynamic sport and every game is a universe in itself. I am not saying the Elo should be done away with completely but just that it is not perfect and evolution is required.

    ReplyDelete
  3. Alexis Cochran, New ZealandFebruary 7, 2011 at 4:10 AM

    what is it all over my head let's just talk of how pretty Alexandra is and how pretty her game is. Who cares about the Elo when you have a lovely Chess Queen before you to discuss. You guys are nuts.

    ReplyDelete

Note: Only a member of this blog may post a comment.