Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: I hate null moves

Author: Tim Foden

Date: 01:12:50 01/30/04

Go up one level in this thread


On January 30, 2004 at 03:40:57, Tord Romstad wrote:

>Recursive null move pruning is an effective way to speed up the search of
>a chess engine, but I really detest its unfortunate side effects.  Most of all
>i hate its tendency to cause search inconsistencies, and that it tends to prune
>away strong lines where a piece must traverse a long and complicated path in
>order to reach a good square.  I try to cure the second problem by using a
>dynamic reduction factor, and I also plan to experiment with Markoff's idea
>of trajectories to see if that can help to reduce the problem further.  I have
>no idea what to do with the search inconsistencies.
>
>Most of all, I would like to throw away null move pruning (at least in its
>traditional form) completely.  Unfortunately my attempts to get rid of it have
>so far been very unsuccessful.  It simply slows down my engine too much, and
>the other pruning I do is not enough to compensate.  I really need some
>effective replacement for recursive null move pruning.
>
>Are there any known algorithms which are close to null move pruning in terms
>of efficiency?  It is clear that such techniques exist, for instance whatever
>Junior and Tiger do appears to be at least as effective.  But I cannot recall
>having seen any really interesting ideas discussed in public.
>
>Tord

When I was in Graz I had a quick look at the books that were on sale.  One of
them had the idea of "Fail High Reductions".  It was in one of the books of
collected papers, possibly of the AGC, but a few years old now.

I scribbled a few quick notes down...

 Use evaluation to deduce threat value t

 In search, if

   eval - t >= beta && alpha + 1 == beta

 then

   reduce depth by 1

... (I said they were quick notes didn't I? :))

It's not something I've yet tried in GLC as GLC doesn't evaluate at each
internal node, and it doesn't have an very good idea of the value of a threat
either, but the paper seemed to think it worked quite well.

Cheers, Tim.



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.