Author: Dana Turnmire
Date: 05:52:14 05/29/01
Here is an article that appeared in the 1996 issue of Computer Chess Reports. Quotation from Thoralf Karlsson's (chairman of SSDF) comments to the 8th rating list: "It seems that the opening book has contributed significantly to the rating increase of MChess Pro 5.0. Sometimes the outcome of the game is already decided when MChess Pro 5.0 starts to think." So, although M-Chess Pro 5.0 may play better chess than his predecessor, it mainly owes his leading position in the Swedish ranking list to his opening book. Looking at the results of M-Chess Pro 5.0, it is particularly well tuned for Hiarcs and Genius. In the long run this booking will lead to inflated ratings. In my opinion, there is only one solution to the problem: They have to turn the opening books off when testing the programs! In order to avoid playing the games from the initial board position, they should establish a set of maybe 100 different starting positions. Of course, these positions ought to be balanced in the way that they equally contain closed/open positions. To avoid the return of booking, these starting positions should be kept secret. Like that, the ratings would again truly reflect the real playing strengths of a given program. You may argue that the opening book is part of the program. But I feel that we've reached the point where the role of the opening book becomes predominant. To me the chess engine itself is a hundred times more important than the opening book. Ratings based on tuned opening books are of NO interest to me. Anyway, the programs' strengths against humans is not affected at all by only adding some killer variations to the opening book. Humans can easily avoid the opening book by leaving the theory in an early stage of the game.
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.