Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A new moveordering technique?

Author: Rémi Coulom

Date: 01:09:27 11/20/01

Go up one level in this thread


On November 19, 2001 at 18:45:26, Erik Robertsson wrote:

>I was fiddeling around with Enhanced Transpositional Cutoffs but what I gained
>nodes I lost i speed...
>
>Anyway I came up with an idea related to ETC but involving move ordering. Just
>wanted to know if anyone has tried this before, and what the results were:
>
>Close to the root of the tree, generate all moves and see if you can find the
>resulting positions in the hashtable. Now, disregarding the search depth, if the
>value of the position according to the hashtable would result in a cutoff with
>regards to the current alpha-beta window, order the move to be searched first.
>
>The idea of this is that if a value of a move from a shallower search would fall
>out of the search window, then maybe the value of the move won't change that
>drastically in a deeper search and is pretty likely to cause a cutoff in the
>current search.
>
>According to my tests, I reduced the tree about 2-3%.
>
>Any opinions?

I do something very similar (although not identical) in The Crazy Bishop. I do
not remember exactly how much it reduces the search tree, but I remember it can
be much more than 2-3% in the endgame.

Rémi



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.