Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: null-move failure detection

Author: Will Singleton

Date: 14:35:20 11/19/98

Go up one level in this thread



On November 19, 1998 at 16:03:58, Peter Fendrich wrote:

>On November 19, 1998 at 15:43:56, Will Singleton wrote:
>
>>All,
>>
>>I am using a null-move error finder that works like this: if a null move causes
>>a beta cutoff, I set a flag instead of returning the null move score.  If the
>>regular search then doesn't cause a cutoff, it's a failure.
>
>I hope I understood this right:
>You are using this function as a "debugging" tool and not in your ordinary
>search engine, or...?

Yes, as a debugging tool.


>>
>>This happens in several of the wac positions, causing problems.  One solution
>>I've heard of is to disallow null moves in the principle variation.  Since I
>>generally ignore the pv during the search, I'd have to do a fair amount of
>>coding to test this.
>>
>>Does restricting null moves in the pv guard against null-move failures?
>
>No it doesn't. A null move error might be the reason why a variation didn't
>became the pv in the first place and that doesn't change with this restriction.

Right, I had just figured this out.


>Another problem with this is that the pv and it's branches take up most of the
>search time. The null move approach is most valuable right here!
>

That's another good point.  btw, I don't do a zugzwang test, just disallow null
moves in the endgame.  How about you?

Will



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.