Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Quiescent Explosion

Author: Anthony Cozzie

Date: 08:21:44 06/27/03

Go up one level in this thread


On June 26, 2003 at 22:45:17, macaroni wrote:

>I recently wrote a computer chess program, using alpha-beta, null moving, and
>quiescent search, in the main search function I use a history heuristic to sort
>moves, and that seems to be doing just fine, I can't say the same for my q
>search, the sorting procedure I use for that, is biggest capture, smallest
>attacker. However, when I do a ply 5 search, i get 23,000 standard search nodes,
>which seems acceptable to me, but I get 180,000 q nodes, which seems ridiculous.
>Is this as bad as I think it is? is it expectded? should I just make my Eval,
>MoveGen, MakePosition and UnmakePosition functions faster (if possible)? Also,
>my program manages 75,380 nodes per second, is this high? someone once told me
>that a high node/sec count is not always good.
>Thanks everyone

look up ernest heinz's article on futility pruning.  I use a slightly more
complicated move ordering, but MVV/LVA + futility pruning will get you to a 1:1
or 1:1.5 Node:QNode ratio.

75knps is not good.  A new program should have a very simplistic eval *and*
relatively bad move ordering and pruning, all of which increase your nps.
Assuming your hardware is reasonable (>1.5G athlon) I would say a program with a
very basic eval should be somewhere around 400-500 kn/s.

anthony



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.