Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: quiescence deep

Author: Volker Böhm

Date: 14:05:31 08/04/04

Go up one level in this thread


On August 04, 2004 at 16:46:51, José Carlos wrote:

>On August 04, 2004 at 16:04:00, Volker Böhm wrote:
>
>>Hi,
>>
>>Spike does worse when cutting from see in quiescense too. Don´t know why, but I
>>haven´t found a bug in see.
>>
>>Greetings Volker
>
>  You both cut only captures with negative SEE score, right? And only in
>quiesce, right? Make sure you don't have a bug, for example, if you don't sort
>captures, and you find a capture with negative SEE, you still need to check the
>other captures... (it's a good idea to sort them according to SEE score).
>  Cutting by SEE migh lose some good captures (for example, a N takes pawn,
>which is defended by another pawn, but discovers a bishop attack against enemy
>queen), but the tree reduction is so big that the overall result is clearly
>positive. At least for me, and I think, for most programs.
>  If you have any doubt with implementation, you can post some code here and
>I'll try to help.
>
>  José C.

Thanks for your proposal, but I think the code does well. I have checked about
hundret positions where SEE with negative score proposed a cutoff in quiescense
where a cutoff should not be made. Second I have a fast not reliable SEE
algorithm implemented totally different and checked diffenrent results of these
two algorithms. I allways found a problem whith the principle of a static
exchange evaluator, not a bug in a code.

Typical problems are:
1. Somhow pinned pieces
2. Pieces defending two fields (you can exchange with a little loss to leave
another piece in another field undefended)

I am currently using SEE for capture sorting only.

Greetings Volker



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.