Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: If you like to solve a mate...(proof number search)

Author: Gian-Carlo Pascutto

Date: 14:24:12 02/06/01

Go up one level in this thread


On February 06, 2001 at 16:55:09, Dan Andersson wrote:

>What kind of proof nuber search do you use? Is it pn or pn^2? I recently
>rekindled my interest for best first algorithms and do think that they will
>yield better results than enhanced ab-search in the near future. I don't know
>if you have tried to program a crazyhouse chess program, but pn search makes an
>excellent leaf node evaluator if used in a normal ab-search.

Huh? I'm very curious as to how you are going to use the pn-search
result to feed any kind of decent evaluation. For suicide and other
mobility-based variants I can believe that, but for crazyhouse or
normal chess, no way. It's horribly expensive.

I've tried pn search on normal chess, crazyhouse, suicide and losers.

For suicide it works perfectly, you don't even need an alphabeta
to get a real evaluation (that is, to get a decent program. if you
want a really good one you _will_ need alphabeta)

For losers it works, but only for finding mates. Basing your play
on it leads to total garbage.

For crazyhouse and normal chess I found it to be a loss overall,
so I no longer use it for those.

If you want to toy with it, Sjeng has code for all of the above,
including pn^2, although parts of it are commented out. If you
know the algorithm you shouldn't have any trouble.

--
GCP



This page took 0.03 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.