Author: David Eppstein
Date: 09:22:21 12/06/00
Has anyone else been following Jeff Sonas' match predictions on Kasparov Chess? e.g. http://www.kasparovchess.com/serve/templates/folders/show.asp?p_docID=13336 As far as I can tell, he is doing something like the following: for each opening position reached sufficiently often by both players, compute the players' performance ratings based on their previous games in that position. The difference in performance ratings gives you an estimate of how likely each player would be to win, in case they happen to reach that position. Then, do your standard minimax (or alpha-beta) search, from the starting position, using these performance rating differences as your evaluation function. The result is a predicted opening choice and a predicted match result (with the assumption of course that players will tend to steer for lines where they perform well and their opponents perform poorly). I think it's an interesting way of using computers for doing chess analysis, but it also seems like maybe a good way for someone who has lots of opponent data (Bob?) to choose book lines by using the knowledge of who the opponent is. Has something like this been tried? If so, how well does it work?
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.