Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Null move reductions

Author: Ernst A. Heinz

Date: 16:31:04 10/04/98

Go up one level in this thread


On October 04, 1998 at 19:06:15, Roberto Waldteufel wrote:
>
>Hi all,
>
>I wonder what reductions various programs use for the null move. I reduce by 2
>plies, but I believe a one-ply reduction may be more usual. However, I have
>found R=2 produces quite good results in my program. I would like to hear of
>others' experiences.

Below I quote again from my article about "How DarkThought Plays Chess" as
published in the ICCA Journal 20(3) and electronically available from the WWW
page of "DarkThought" at URL <http://wwwipd.ira.uka.de/Tichy/DarkThought/>.

*******************************************************************************

Search Parameterization

[...]

Null-Move Search.
      Generally regarded as both a curse and a blessing of
      microcomputer-based chess programs, the realm of null-move
      pruning provides ample opportunities for many improvements (e.g. by
      parameter fine-tuning). In DARKTHOUGHT, null moves may be
      layered (i.e. disabled in either the upper or lower parts of the search
      tree up to a predefined depth threshold), made non-recursive or even
      switched off completely. The first DARKTHOUGHT delayed null moves
      until depths  <= 5 after Don Dailey (of SOCRATES and STARSOCRATES
      fame) enthused about this scheme during the 8th WCCC in Hong
      Kong, 1995.

      Like many other chess programs, DARKTHOUGHT currently uses a
      depth-reduction factor of 2 for null-move searches. But other values,
      especially depth-adaptive ones in combination with various material
      thresholds for disabling the null move in endgames, are constantly
      experimented with. The same holds for depth and score margin of the
      deep-search extension which default to 2 and the value of 3 Pawns
      respectively.

[...]

*******************************************************************************

=Ernst=



This page took 0.1 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.