Computer Chess Club Archives




Subject: Re: Can nullmoves behave like this?

Author: Severi Salminen

Date: 03:15:47 01/10/01

Go up one level in this thread

>I do not have this weird fail high fail low problem a lot of times.
>I had it when i forward pruned bigtime in DIEP or when i didn't store
>extensions in hashtable.

This particular position is the only occurence of this phenomenon I have seen.

>If you do things like threat or mate extensions you must store it in
>hashtable or you get the weird fail high fail low because of
>inconsequent way of searching!

Ok, but I don't use hashes and only extend on checks now.

>Perhaps an idea is to simply try to search without nullmove the last 3 ply
>and see what that gives for kind of effects?

I'll try, but now I'm in another town than my engine :(

>So keeping nullmove on the top plies?
>Another interesting possibility: perhaps your repetition keeps on working
>after nullmove. Is nullmove implemented such that you don't try to
>get repetition from positions before nullmove?

Yes, I counted this in. But as said, my progarm behave odd on this position even
after removing repetition check. What basically is happening is that with a
small aspir. window nullmove fails high at ply 2 (so white plays Kb3 and then
again) but with infinite window this doesn't happen. And if I recall right
futility pruning didn't affect to this. And I don't want to trace where this
"oddity" happens (a 8 ply search), too much debugging...If this is "normal" I
can live with it, but if it is a bug - damn!


This page took 0.35 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.