Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Search instabilities when re-using hashtables

Author: Robert Hyatt

Date: 16:11:30 06/17/02

Go up one level in this thread


On June 17, 2002 at 12:39:30, Gian-Carlo Pascutto wrote:

>On June 17, 2002 at 11:25:36, Robert Hyatt wrote:
>
>>Your first move returns a score of +1.  Your next move fails high.  With the
>>wide window, it then returns a score of -1.  You stick with that.  Do you
>>really want to?  Most likely it is correct to do so, but you just _lowered_
>>the score at the root.  Will you now use a lot more time trying to fix a
>>problem that isn't there?
>
>Here you're talking about the PVS searches. I'm having the aspiration
>window search in mind.
>
>If this happens in a PVS research, I ignore the fail high.
>
>You can't do that (as easily) in the aspiration search.
>
>--
>GCP


I agree, but I was talking about the aspiration window search at the
root.  You get a fail high based on an old hash table entry.  The
research can fail low if you use the new bound of beta, + infinity.

If you try to fix that by searching with original value, + infinity, you can
get a _lower_ score and confuse things.  I don't see any way to solve that.



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.