Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: What kinds of (forward) prunning are known?

Author: Ernst A. Heinz

Date: 10:08:55 05/06/00

Go up one level in this thread


Hi Jan,

>I would like to know what kinds of pruning are known (and/or used). To be
>correct, I don't mean alfa-beta (which is safe), but I mean rather kinds of
>forward-pruning(FP) (ie: not always safe...). I use this one: when it's last
>ply to horizon AND  eval()>=beta AND not in check THEN return eval(). Are there
>several (probably more brutal (?)) techniques in use (except null-move)?

I think the technique that you describe is already _VERY_ brutal
and not reliable at all. Looks like an "extended stand-pat cutoff"
to me which is widely known to not work well at all ...

In my book "Scalable Search in Computer Chess"
(http://supertech.lcs.mit.edu/~heinz/node1.html), I distinguish
between dynamic and static forward-pruning schemes. Null-move
pruning is the most successful dynamic one so far, while the
various flavours of futility pruning (normal, extended, in the
quiescence search) are the best static ones known today. The
book and my earlier articles in the ICCA Journal elaborate on
many of these techniques.

Please enjoy,

=Ernst=



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.