Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: about Falcon-Crafty test positions(bug in yace?)

Author: Dieter Buerssner

Date: 20:07:34 07/17/04

Go up one level in this thread


On July 16, 2004 at 14:05:48, Uri Blass wrote:

>I heard that Ra8 is very hard move to find but it seems to me not very hard for
>Yace that can avoid it on A3000(512 mbytes hash) in 130 seconds.
>
>Second mistake seems even more easy to avoid but surprisingly it seems that
>yace0.99.87 has antisymmetric evaluation because if I give it the position
>after the second mistake it has problem to recognize a draw score and see
>advantage for white(I see the same thing even when I delete learning files so it
>is not a result of previous learning)

Thanks Uri, for pointing this out. I can reproduce it here (from the positions
alone even, without the game history). I have not found the problem yet - it
doesn't look trivial. I fear, I have to investigate a dump of the search tree up
to depth 11 from the last position you gave. Such a dump is very huge, and
tedious to investigate. At the moment I am thinking of some method to only
include critical lines into the dump.

Regards,
Dieter



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.