Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Nullmove question

Author: Vincent Diepeveen

Date: 12:53:25 11/05/01

Go up one level in this thread


On November 05, 2001 at 15:12:13, Gerd Isenberg wrote:

>On November 05, 2001 at 14:34:43, Robert Hyatt wrote:
>
>>On November 05, 2001 at 13:02:31, Gerd Isenberg wrote:
>>
>>>why not
>>>
>>>val = -AlphaBeta(depth - 1 - R, -beta, -alpha);
>>>
>>>
>>>Gerd
>>
>>
>>In PVS that would be find.  But the question is, what do you want to
>>know from the null-move search.  For a fail high, all you can discover
>>is "if I do nothing, the score comes back >= beta" which is enough to
>>convince me that I can simply return beta here, as playing _any_ move
>>should raise the score even higher (non-zug of course).  Does it matter
>>_how_ good the score is?  Or just that it is >= beta?
>
>In most nodes in PVS there is already a null window, but using the full window
>here seems to be better for my program, empirically determined with BT2630 and
>LCTII - may be some true scores in the transposition table? In the conditional
>verification search i have slighly better results with an explicite null window.
>No idea about the influence in practical games.

for diep nodes reduction happens if i use [ beta-1 ; beta ]
versus [alfa ; beta ]. This for the reason like Bob described.





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.