Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To check or not to check, this is the quiescence question

Author: Robert Hyatt

Date: 13:47:51 10/13/03

Go up one level in this thread


On October 13, 2003 at 00:28:11, Charles Roberson wrote:

>
> Hi Omid,
>
>    I have been experiementing with the same. I generate checks but do not
>generate all responses to checks (that made things worse). Adding checks did
>make test scores go up noticably. I am still analyzing the impact on actual game
>play. There was an interesting article in ICCA on this subject as well about a
>"strategic" quiesce routine.
>
> Charles


If you don't do all responses to checks, how can you follow checks?  IE
what if there is only one way out, and you don't follow it.  Do you use
the resulting (and wrong) mate-in-N score?  What if there are two ways
out, one that draws, one that loses a knight.  What do you do if you fail
to follow the one that draws and assume you lose a knight?

You should not do all evasions if the moving side has any stand-pat opportunity
of course, as he would always choose to stand pat rather than getting mated,
which would be a wrong score backed up.  But if you are at an even ply, and
you were in check at _every_ q-search node for even plies, then forcing all
replies here makes good sense as if you find it is mate, it is forced.  It
won't necessarily be the shortest mate, but it will definitely be mate.




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.