Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: move ordering and node count

Author: Steve Maughan

Date: 09:24:42 03/29/04

Go up one level in this thread


After hacking around in Excel for a couple of minutes I came up with this
formula.  Of course it's an engineers fomula and not a mathematicials formula
i.e. it looks about right and I'm not really interested in the proof :-))

Node = 2 * sqrt(moves ^ depth) * sqrt((1 / MoveOrdering) ^ depth)

It's based on the thought that when the MoveOrdering stat falls we need to
search more than one move at a theoretical cuttoff node i.e. alternative depths.
 Also the effect is going to be multiplicative and as the MoveOrdering
approaches 1 the nodes approaches the theoretical minimum.

Thoughts?

Interestingly this means that the difference between a move ordering stat of 0.9
and 0.95 for a 10 ply search is 31% more nodes.  At a 15 ply search this
increases to 50% more nodes.  So improving the move ordering has more effect at
slower time controls - I think we knew this anyway.

Regards,

Steve



This page took 0.01 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.