Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Alpha-Beta fool: for Ed Schröder

Author: Uri Blass

Date: 00:11:49 03/28/01

Go up one level in this thread


On March 28, 2001 at 01:14:02, Ed Schröder wrote:

>On March 27, 2001 at 17:56:15, J. Wesley Cleveland wrote:
>
>>Ed Schröder posted a few weeks ago about experimenting with a larger alpha-beta
>>window at low plies to improve move ordering at greater depths. I never saw
>>anything more on this. Did I miss it ?
>
>I remember to have something posted about that some months ago. The trick
>is not a wider alpha-beta window but to fool beta with some added fixed value
>that forces a normal cut-off not to be taken, thus oblige the search to go for
>the next move. The goal is to get a better move-ordering at deeper iterations.
>
>You can only do this in the early iterations, say the fisrt 5-7 iterations. I
>have tried all kind of variations but never found a formula that was clearly
>superior to Rebel's current move ordering.
>
>Still I think there must be something to gain from the idea.
>
>Ed

I remember from your posts that you found a way to do Rebel 1% faster by this
idea and I guess that at longer time control(or with faster hardware) it may be
more than it.

Uri



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.