Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Evaluation Should Be Winning Probability - Not Pawns

Author: Jeroen van Dorp

Date: 07:58:05 01/16/02

Go up one level in this thread


It seems that Graham feels most uncomfortable with the absolute value that is
suggested when you evaluate on pawn difference calculations.
The win/lose chance hasn't got that problem, as it suggests no absoluteness, but
chances.

A problem, because you only can see chances over a longer period.

But if you take chess engines in account, we only want one thing: a perfect
chess engine. A perfect chess engine knows the solution to chess, or "is able to
solve chess". In that case it can tell -1, 0 or +1 as evaluation: You will lose
(mate in ##), We will draw, or you will win (mate in ##).

As that's not possible yet for more than 5 or 6 pieces, the engine use a
modified assessment based on pawn value difference. It gives more information
than a rating system like win/lose chances, but -and if Graham thinks so- in
that I agree with him that chess engines suggest an accuracy that's not there.
Yet. :)

J.



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.