Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Static Exchange Evaluation (SEE) for pruning in quiescence (?)

Author: Bo Persson

Date: 04:36:37 08/20/03

Go up one level in this thread


On August 19, 2003 at 08:07:07, Uri Blass wrote:

>On August 19, 2003 at 08:05:39, Uri Blass wrote:
>
>>On August 19, 2003 at 07:43:08, Bo Persson wrote:
>>
>>>On August 19, 2003 at 06:19:59, Joost Buijs wrote:
>>>
>>>>
>>>>To me it looks wrong to skip losing captures from the quiescence search if the
>>>>losing captures are determined by a SEE that is wrong in some cases, e.g. pinned
>>>>pieces. If you use the SEE for move ordering purposes only this problem doesn't
>>>>exist.
>>>
>>>But it is always only an approximation, you just want it to be good enough to
>>>avoid flat out blunders.
>>>
>>>In a position like this
>>>
>>>[D]8/1p6/r1r5/8/8/8/8/R1R5 w - - 0 1
>>>
>>>SEE will believe that the black rooks are both defended, but they are not. This
>>>is not a "pin", but an "overload" which is still missed. You just can't get it
>>>all (cheap).
>>>
>>>
>>>Bo Persson
>>
>>They are defended.
>>
>>Rxa6 Rxa6
>>
>>if you want a better example then it is better to put a pawn at b6 and not only
>>at b7.
>>
>>Uri
>
>It is still wrong because Rxa6 Rxc1
>
>It is better to put 2 knights and not 2 rooks to show that SEE may miss a good
>capture by white.

Yes, you are right. My point was that if SEE is used to resolve captures on a
specific square, you miss the fact that a single piece can be involved in
several of these "shoot outs". In real play it often cannot, at least not
simultaneously.

That another piece might be pinned doesn't make that much of a difference.

Bo Persson


>
>Uri



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.