Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Aspiration search

Author: Dave Gomboc

Date: 09:46:17 08/13/99

Go up one level in this thread


On August 13, 1999 at 10:22:59, Ulrich Tuerke wrote:

>On August 13, 1999 at 02:19:30, Scott Gasch wrote:
>
>>Hey,
>>
>>I have a (stupid?) question about aspiration search.  I understand the concept
>>is to atrificially limit the search window in an effort to prune faster.
>>However, a value returned from Search(pos, a, b) will always be within the a-b
>>window.  I.e. a <= value <= b.  How, then, does one tell if the value returned
>>is bogus and needs to be re-Searched or whether it is good?  Which does a value
>>== a or value == b mean?
>>
>>Thanks,
>>Scott
>
>Usually, the search window is defined without the boundary, i.e.
>a < value < b   is an exact score; a = value or b = value means that you have
>to proceed a verification search.
>
>BTW, the alpha beta search of my program returns also values far outside the
>alpha-beta region; these can be taken as bounds for the verification search.
>I think that I remember vaguely that there are 2 kinds of alpha-beta
>implementations: "fail soft" and "fail hard". I think that "fail soft" always
>returns a value a <= value <= b ?
>It seems that you are using "fail soft" whereas I use "fail hard".
>
>I am sure that Bob Hyatt can help.
>
>Regards, Uli

It's the other way: "fail hard" means that the score will be held within the
range [a,b], but and "fail soft" will return a score outside the range [a,b] on
occasion.

Dave



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.