Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A story about NPS

Author: Severi Salminen

Date: 09:49:15 01/26/01

Go up one level in this thread



>What you say is extremely to the point. I have had similar results with Chess
>Tiger. I can easily increase my NPS and beat everything else on NPS (on PC, I
>mean), but my program would be significantly weaker.
>
>The kind of NPS increase I'm talking about is not by removing chess knowledge. I
>can keep exactly the same evaluation, and by changing several things in the
>search (removing pruning algorithms), I can increase my NPS dramatically. But
>the program will be weaker, even if it is 5x faster in NPS.

Yes. What made me worry is that when I removed SEE from my engine and played
(just a couple of) games between a version with SEE there were no differences in
tactical play. Could it really be that SEE slows things so much down that the
benefit isn't really significant? I hope this is not the case because I spent _a
lot_ time adding those xray attacks in SEE...

In Qsearch I stop searching more moves in that node when
eval(node)+move[i].see_score+PAWN_VALUE drops below alpha (they are sorted).
Could this be even more selective? Maybe the PAWN_VALUE could be 0? Have to
try... Maybe it would make things brighter for SEE. (I might remember something
wrong as I don't have access to source code right now.)

Severi



This page took 0.05 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.