Author: Serge Desmarais
Date: 18:54:47 08/27/98
Go up one level in this thread
On August 25, 1998 at 02:43:59, Shaun Graham wrote: > Substantive way to judge program strength improvement other than ELO? I ask >this question, because most programs coming out will only be 10-20 elo points >stronger than previous versions. Since this is the case we could achieve a >similar gain in ELO by just creating a more highly optimized opening book. >Further since most new programs always claim to have an improved opening book, >how can we know there has been any real improvement? I thought that perhaps you >could do the no opening book test, but i'm not certain that doing so shows much, >because prog A might play the middle game and endgame great, but sucks in the >opening. further it's not realistic in the sense, that all strong >players(humans) actually do have stored mental opening books. That is why there exist test suites, like the Louguet's. But since these tests are public, it is theoretically to fine tune a program to score well in these tests while it would be weak in other similar but different posiyion (?). The Nunn's match is another way, since both programs alternatively play White and Black, starting from a certain known book position, but without using their own books... As what you said about the human players and their mental opening book, they aren't stuck to play it all the time. I remember an Anand's comment about one of his game against Lautier (?), I think. He reported having analyzed the position of the game very deep at home before the tournament (in preparation against Lautier?) but while he was playing Lautier, he decided to completely ignore the rest of his deep analysis while in the middle of it and he played (after mental calculations) a move he never considered at home. After the game, he was convinced that the move picked up during the game was better than the one analyzed for hours at home! Computer cannot do that. Other players are/were known to often improvise durina a game : Alekhine, Larsen, Liuboyevitch etc. while others stick all the time to their preparation (Kasparov and Karpov are an example). Serge Desmarais
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.