Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Null Move Pruning and Draw Detection

Author: scott farrell

Date: 05:27:32 01/15/03

Go up one level in this thread


On January 15, 2003 at 07:42:14, David Rasmussen wrote:

>On January 15, 2003 at 07:35:38, Uri Blass wrote:
>
>>
>>I do not see how null move pruning effect stalemate detections.
>>
>>Cases when my evaluation detects stalemate immediately after null move are very
>>rare and I do not see how these cases can do demage to the program.
>>
>
>I'm not saying that any of these things "damage" the program, I just want to be
>absolutely clear *why* not, if not.
>
>In normal situations the logic of null move pruning makes sense:
>"If I do nothing, how good can my opponent do?"
>But in the presence of repetitions, fifty moves rule and stalemate, it becomes
>kind of pathologic, I think.
>"If I do nothing, he/I is/am stalemate" or
>"If I do nothing, he/I will reach the fifty moves rule"
>"If I do nothing, there I/he can force a repetition"
>All of this makes no sense, since a null move isn't legal.
>
>/David

I think you have it slightly mis worded.

It should sound more like this:

If I dont do anything about it, things will probably deteriate to draw.

This might be a fail-high, which means the move is too good, and the opponent
wont let things get here in the tree, lets just fail high righ now.

If it isnt a failhigh, it was a waste of time. So maybe if it is nearing a draw,
and beta is less than a draw, the null move might help. If it is nearing a draw,
and a drawscore wont fail high, then you'll just be wasting nodes/time.

Obviously, you have to be really careful when checking for draw by rep as you go
through the nullmove barrier.

Scott



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.