Computer Chess Club Archives




Subject: Re: fail-high on first move stat (86%)

Author: Scott Gasch

Date: 09:07:55 07/29/04

Go up one level in this thread

On July 29, 2004 at 03:37:11, Michael Henderson wrote:

>I'm getting an 86% average success rate for this stat in mid-game.  I'm thinking
>this is too low since most programs have 92-94% and the diff. in % is
>Here is my move ordering scheme:
>1. hash table move (48MB)
>2. winning captures, even captures
>3. IID move
>4. killer moves (2 slots)
>5. rest of moves based on history, losing captures
>eval is nothing but piece-sq tables.  Maybe I need special root-move ordering?

Every time your program fails high and it's not the first move, dump a list of
moves generated and tried so far.  This gets really noisy so redirect your
output to a text file.  Then go through it by hand and see what your engine is
doing.  If you have never done this before it should give you all kinds of ideas
about how to improve your move ordering.

Root move ordering is not the problem if over the whole tree you are at 86%.  I
suspect you have some subtile bug in your move scoring.

Good luck,

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.