Author: Tom Kerrigan
Date: 03:48:05 08/12/98
Go up one level in this thread
While this sounds good, it's not exactly practical. A program that checks every move in its book will take out all the gambit lines, although some of the lines may be good and it may be important to have them when opponents play gambits. A program that learns by analyzing its own games sounds good too, but nobody knows how to do this effectively. -Tom On August 11, 1998 at 22:01:54, blass uri wrote: >Some programs like fritz5 with powerbook, Mchess7, and chesssystemtal >have big books. >The result is that they can follow wrong lines. >For example Fritz5 and Mchess7 can do a draw in the opening with white > >I think every move in a book the program play without checking should be checked >by the program before the game. >If the programs check 100 positions per day they can have in 3 years >a book theory of 109500 positions and I think it is enough. >I think grandmaster do not remember more theory. >I think a good idea for programs is to concentrate in a small number of lines >they understand. > > >They can have a big knowledge book but they should not use it without thinking. >If they lose they can learn by analyzing the games(1 move per hour in the moves >before the opponent has the advantage) and see where they did mistakes. > >There should be time of at least 12 hours between the games. > >Another option is to do singular extentions to the line of the game in the next >game. > >Uri
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.