Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Null move and move ordering stats

Author: Tony Werten

Date: 01:06:53 12/20/01

Go up one level in this thread


On December 20, 2001 at 03:30:13, Severi Salminen wrote:

>>Either you have extremely good move ordering, you have a bug, or you're counting
>>in quiescence search also, which is "cheating". 96% is very very high. Crafty
>>gets 90% here.
>
>I also tested with Crafty and did a search to same depth than on Requiem and got
>worse percentage and wondered waht causes the difference. You were right, I
>_was_ counting in qsearch also (I had thought it was the standard), BUT: when I
>removed the calculation in qsearch the figure increased by 0.2 (95.9% -> 96.1%).
>And without null moves it was still 98.9%. So I guess I have an "extremely good
>move ordering" :) But it was only this position, on other positions the figure
>might be lower (or highr on the other hand).
>
>This is how I calculate (this is done, if score>=beta):
>
>FailHighCount++;
>if(moves_searched==1)
>  FirstMoveFail++;
>
>This is not done in rootnode, so it might bias the results ;)

I don't want to sound unfriendly, but I think you forget to increase
moves_searched in your search. 98.9% is impossible without hashtables. (even
with, it's doubtfull) You must have a perfect SEE to get this. Even then, the
amount of noncapturemoves that are best is higher than 1.1%

Tony
>
>Severi



This page took 0.01 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.