Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: reducing null-move searches

Author: Robert Hyatt

Date: 07:17:19 11/17/97

Go up one level in this thread


On November 17, 1997 at 08:10:51, Alessandro Damiani wrote:

>Hello!
>
>Before searching the null-move or doing any form of forward pruning, I
>think it should be faster if we test whether evaluate(position)>=beta.
>Since I own a slow machine (MC68020) I would like to know how it works
>with deep search (in the middle game). Any comments?
>
>Bye
>
>Alessandro

would be very dangerous.  I assume you mean something like this:

   if (evaluate() >= beta) return(beta);
    ..  do null-move search here  ...

The problem with this is that a null-move search is more conservative,
in that *at least* we get to the capture search to see if something is
hanging badly.  The above would have *severe* horizon effects.  IE
imagine a case where my opponent sacrificed a rook for a mating attack.
I'm a rook up, so the above test would succeed and I'd quit and not
notice
I'm getting mated, only that I am a rook ahead.  But the roof would fall
in
later of course...



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.