Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: fixed depth evaluation problem

Author: Russell Reagan

Date: 15:53:15 06/16/02

Go up one level in this thread


Check out Bruce Moreland's webpage about alpha-beta and make sure that you have
implemented it correctly. As another poster mentioned you could have a sign
switched or something. His webapge is http://www.seanet.com/~brucemo

The other thing that come to mind is that the weights in your evaluation
function might be weighted incorrectly. You mentioned the throwing of pawns from
in front of the king instead of castling or something like that. The way I
understand this is in the following example.

[D]r1b2rk1/ppp2ppp/2pn1n2/8/8/5Q2/PPP2PPP/R3K2R w KQ - 0 1

For example, castling in this position is a good idea for white. Fritz thinks
white should castle. Now, you might have some odd thing in your evaluation
function that would make your program want to play g4 to prevent Bf5 which
attacks the unguarded pawn on c2. So perhaps that "thing" in your evaluation
function is given more weight than the importance of king safety, so your
program is chasing after this weird thing instead of protecting the king.

It sounds more like it's something in the search though.

Russell



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.