Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka and a different evaluation

Author: Christopher Conkie

Date: 04:09:27 02/14/06

Go up one level in this thread


On February 14, 2006 at 06:31:34, Thorsten Czub wrote:

>the evaluation directs the search. search without good evaluation computes many
>senseless moves.

Yes, I agree fully with this. You see things like this......

#define	VALUE_PAWN      100
#define	VALUE_KNIGHT    300
#define	VALUE_BISHOP    300
#define	VALUE_ROOK      500
#define	VALUE_QUEEN     900

......in chess engines. A very static evaluation of piece values. Surely these
values are dynamic as the game progresses.

I also see many piece square tables constructed to help but they are also very
static. The amount of moves available and the range of the piece in question, in
the given position, and what kind of motifs/strategies that piece can create
from where it sits determine the actual value of a piece.

Statistically in human play you can see this dynamicism.

I think we all agree that pieces can be worth nothing in chess because of where
they are in a certain position, in fact they can actually hinder. Pieces can be
worth nothing or less in chess, so why are they always worth something positive
in chess engines?

A knight on the rim (of value 300) is dim for the entire duration of a game.

The point being that it could be worth -300. This realisation made on a position
by position basis is surely superior.

Christopher




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.