Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A Question on simple Alpha-Beta versus PVS/Negascout

Author: Ernst A. Heinz

Date: 11:19:27 03/22/00

Go up one level in this thread


Hi Bruce,

>Some of these algorithms attempt to return valid scores that are outside the
>window.  I've never figured out how to do this.  One of the systems that tries
>to do this doesn't work with a variable depth search.

Why does fail-soft PVS not work with variable-depth searches
in your opinion?

I prefer fail-soft over fail-hard by far and it works perfectly
well in "DarkThought" (http://supertech.lcs.mit.edu/~heinz/dt/).

The initial chapter of my book "Scalable Search in Computer Chess"
(http://supertech.lcs.mit.edu/~heinz/node1.html) explains the
nature and benefits of the fail-soft formulation on pages 14-17:

  * more potential fail-high cutoffs and

  * possibility of better lower bounds for the researches.

=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.