Author: Thorsten Czub
Date: 17:19:09 10/22/99
Go up one level in this thread
On October 22, 1999 at 12:00:57, Tord Romstad wrote: >The flaw of your car anology is that the evaluation of chess programs are >_never_ correct, unless they find a forced mate or draw. That a chess >programs evaluation function does not change after making a move does not >mean that the evaluation is anywhere close to the truth! Your own positional >understanding, Thorsten, is probably vastly superior to all the chess >programs you own. Unless the program finds a forced mate or win of material, >you should _never_ trust the evaluation. And when the program finds a line >which wins material, you should _always_ check the resulting positions >manually and consider if there is enough compensation for the material. > >If chess programs came with an option to evaluate material only, I would >probably switch this option on when using the program for analysis (to >optimise the tactical speed of the program). > >Tord Again : i never say THAT I AM USING THEM THIS WAY. but i want to say: as long as they work this way, they will not make much progress. because IMO this way is wrong. programs should try to understand about chess. if they don't evaluate resulting in a nice plan, that leads maybe in a win, it is senseless IMO.
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.