Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: perfect ordering

Author: Vincent Diepeveen

Date: 10:04:58 01/09/02

Go up one level in this thread


On January 09, 2002 at 12:50:12, David Hanley wrote:

>I have seen it claimed somewhere that with perfect move ordering, an eight ply
>search would only consume a thousand nodes or so, even only alphabeta ( no
>hashing or forward pruning ).
>
>Is this so?
>
>dave

all chessprograms use nowadays nullmove. this means that
you can do with very little nodes, depending upon evaluation.

Suppose your evaluation is always returning 0, which makes ordering
near perfect of course. suppose we order such that not a single
extension gets triggered:

 h3 h6 g3 g6 f3 f6 e3 now we have to look at all nodes behind here.

 but for example R=3 at depth 2 how many nodes do we need to
 get a cutoff for other moves than h6?

 Let's give one example:

 h3 g6 nullmove <all moves> qsearch
  1 2   3       4+R

So for all moves at the second move we only need to search number
of semi legal moves there for each possibility at 2 ply.

that's like 20 posibilities x 30 = 600 nodes.

In short real little to refute all moves at root+1 ply.

This where without nullmove total number of nodes needed is gigantic
compared to what nullmove needs.





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.