Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Expanding window

Author: Matthias Gemuh

Date: 09:45:49 07/02/03

Go up one level in this thread


On July 02, 2003 at 11:39:18, Vladimir Medvedev wrote:

>It is written on Ed's page:
>
>"Whenever REBEL gets a fail-low or fail-high at the root position (we are back
>at ply one), REBEL doesn't immediately opens the entire alpha or beta window,
>instead of that it tries a new window of 2.00 and only then when REBEL is faced
>again with a fail it opens the complete (corresponding) window. The 2.00 window
>will do for most of the positions when a fail occurs, that's the tought behind.
>"
>
>Why does he expand window only in the root node?
>Won't this technique work in inner nodes as well? What is the critical
>difference between position at ply=0 and ply=1,2,3 ?


At root, he must have an exact score  (alpha < score < beta) to have a move.
In tree, he prefers a cut-off (score > beta) and no move needed.

/Matthias.



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.