Computer Chess Club Archives




Subject: Re: MTD(f) (was Re: New(?) search idea.)

Author: Robert Hyatt

Date: 06:33:26 01/23/98

Go up one level in this thread

On January 23, 1998 at 02:38:52, Bruce Moreland wrote:

>On January 23, 1998 at 00:18:18, Don Dailey wrote:
>>2. You mentioned extension based on the target score (alpha/beta.)  We
>>used to do this and eventually threw these out long ago because they
>>do cause problems whether you use mtd or not.  So far our program
>>has met your requirements for MTD candidacy!
>Isn't null-move forward pruning an example of this nasty alpha-based

yes...  but we never have problems with a null-move search failing high
in one case and low in another, only because alpha/beta changes.




I have *enough* trouble with a position that fails high, the hash
that it fails high, and then later I look this position up and get a
high because of the hash table.  Then I relax beta, re-search, now the
hash entry can't make me fail high, the search can not see deeply enough
to figure out why this failed high, so I fail low.


This page took 0 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.