Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: revolution in computer chess

Author: Tord Romstad

Date: 11:58:57 01/03/06

Go up one level in this thread


On January 03, 2006 at 14:40:36, Tom Likens wrote:

>On January 03, 2006 at 14:08:26, Tord Romstad wrote:
>
>>Glaurung Mainz does no real pruning, but my latest development version
>>uses a very primitive form of static null move pruning in the last two plies.
>>This is a very recent addition, implemented one of the last few days before
>>Christmas.  During the holiday, I played my longest ever self-play match:
>>A 1920 game match played from the 960 different FRC starting positions.
>>The version with static null move pruning won by 1034-886 against a
>>similar version without static null move pruning.
>
>What time control control?

Just 2+1, and on a rather slow machine (G5 1.8 GHz).

I have also compared the two versions in Noomen matches
against Ruffian 2.1.0 at 4+1 time control, with a slightly
better result for the version with static null move pruning
(56-44 with SNMP, 49-51 without).

>Also when you mention the last two plies, I'm
>assuming you're talking about the last two plies above the horizon (i.e. the
>interface between the normal search and the quiescence search).

Yes.

>I'm actually toying with the idea of changing the type of pruning/extensions
>the engine does based on the amount of time left in the game.  If both players
>are short of time the engine may be able to get away with more agressive
>reductions, while conversely it may make more sense to be conservative if
>there is a large amount of time left.

This could be worth trying, but it sounds like it would require a lot of
tuning to make it work well.

>I probably should know this, but what are exactly are you calling
>*static null-move pruning* ?

Null move pruning without a search.  When the static evaluation
is above beta and the position is sufficiently quiescent, I return
beta.  My current definition of "sufficiently quiescent" is extremely
crude.  I just check that the side to move has no hanging pieces
with value big enough to bring the score below beta, and that
the other side has no pawns on the 7th rank.  I don't consider
pins, checks or other factors at all.  There are clearly lots of
things to improve here.  I am surprised that the results with
my current implementation are not catastrophically bad.

>Yes, the interaction between programmers is the attraction for me as
>well.  I appreciate the effort put out by the various operators, but the
>exchanging of ideas with other programmers is the real draw.  I wouldn't
>underestimate the drama though.  I participated two years ago and it was
>fairly exciting, (of course this statement qualifies me for nerd of the
>year, but I'm afraid this hobby already has done that!)

I am sure I would feel the same way, but only if I were there
to operate myself.  I wouldn't get much sense of drama by just
reading the results the next morning.  :-)

Tord



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.