Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: PV search question

Author: Robert Hyatt

Date: 11:06:35 10/04/01

Go up one level in this thread


On October 04, 2001 at 10:30:59, martin fierz wrote:

>hi,
>
>i once read about a search scheme where you do your normal search to depth X,
>and once you are done with it, you move down the PV 4ply, and do a search to
>depth X-2. since this is a much smaller tree, you get done with it quickly, and
>somehow get an indication of the stability of your search.
>question: is anybody doing this? does it help? and what are you supposed to do
>when you see that your PV search in the end gives a terrible value?
>
>cheers
>  martin


That was Richard Greenblatt's idea.  It was primarily intended to eliminate
the horizon effect.  The idea was that if the new score was better, you ignored
it, but if it was worse, you used that instead of the original score.

the q-search and tactical extensions are _far_ better.  I doubt anybody uses
this today.  It was a mid-1960's approach when programs were searching a
few nodes per second max...



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.