Author: Christophe Theron
Date: 18:59:44 11/24/01
Go up one level in this thread
On November 24, 2001 at 07:33:10, Dieter Buerssner wrote:
>On November 23, 2001 at 08:30:08, Steve Maughan wrote:
>
>>I came across the YACE web site the other day (http://home1.stofanet.dk/Moq/).
>>Interestingly it explains some of the search techniques employed by YACE.
>>
>>One that I haven't come across before is explained as follows:
>>
>>"ALL_MOVES_IN_CHECK: When set, and a capture checks the opponent, essentially
>>one non capture search will be done. This will need a bit more time, but will
>>find mates earlier."
>>
>>Any idea what's meant? My guess is that when a capturing move also gives check
>>extend the search.
>
>More or less, yes. But as GCP has pointed out, this is not really and extension
>(but similar). The idea of quiescent search is to reach a quiet position. This
>is usually defined, as giving the player at move the choice to capture or not.
>But one thing does not really fit inside here: when we are in check already,
>this is not really quiet. But in deep qsearch, doing a lot of things may also
>hurt the performance.
That's something that has already worried me: you do it when your King is
threatened, but not when your Queen is threatened. What's the logic here?
The only answer that I have found is that if you do it when your Queen is
threatened, the search tree explodes!
I don't like my answer because it has little to do with logic, but there is
nothing I can do about it...
Christophe
>Anyway, by default, when the score looks promising (> alpha), and the side to
>move is in check, Yace will not allow to return the positional evaluation, but
>tries to do something about it. The default beheviour here again is to do an
>almost normal search, after which a new qsearch is started. The
>"ALL_MOVES_IN_CHECK" describes this. To me it is not clear, which of the
>different stratagies works best, but the defaul values seem to be reasonable.
>
>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.