Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why don't we use stand-pat score in search?

Author: Mathieu Pagé

Date: 07:25:14 05/05/05

Go up one level in this thread


On May 05, 2005 at 07:51:33, Ricardo Gibert wrote:

>On May 05, 2005 at 07:31:57, Kevin K wrote:
>
>>Do I have to use stand-pat score only in qsearch?
>>If I use stand-pat score instead of null move in search, what will happen?
>>I will return if stand-pat score >= beta.
>
>Your idea will maximize horizon effect errors. It would mirror the apocryphal
>strategy of the ostrich which whenever something bad seems to about to happen,
>it buries its head in the sand to improve its situation.
>
>Not a recommended strategy.

I entirely agree with you Kevin, but for Ricardo, I'll explain why stand-pat
work in qsearch and does not in regular search.

The idea behind stand-pat evaluation is that at a given node in qsearch if all
the possible captures are bad for the side to move then the player would play a
non-capture move. However in qsearch we do _not_ evaluate thoses moves. So we
know the player would not play one of the captures but we don't know wich other
move it would play, so we return the current evaluation (That is the best
approximation of the real value we can do).

During full width search however, all move are examined. So, if they are all bad
moves we have to return a bad score because there is no other possibly best move
(That was the case in qsearch)

Mathieu P.



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.