Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: History pruning

Author: Robert Hyatt

Date: 10:44:48 02/27/06

Go up one level in this thread


On February 27, 2006 at 13:38:33, Charles Roberson wrote:

>
>  A very clear explaniton. Good work.
>  I think that several of us have thought of ways to reduce "late moves".
>  I tried with some static attempts years ago -- they failed
>    due to being to blind to dynamics and to bad initial move ordering.
>
>  If people implement this with say the history technique and sufficient
>   restraints then poor performance might indicate poor move ordering?

Maybe not.  It could be a poor history implementation.  You depend on big
numbers to say "this move should not be reduced."  If that doesn't happen, you
overlook things right and left.  So the "threshold" to stop reducing is an
issue.  The way you adjust the history value is an issue.  How you "age" the
history values (you have to somehow keep them from becoming large and staying
there, or otherwise as the game progresses you end up reducing nothing since
every history value eventually climbs over your threshold to stop reductions.

It takes some experimentation.  A lot.



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.