Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Distance-to-mate vs Best-move tables

Author: Angrim

Date: 15:49:56 10/24/01

Go up one level in this thread


On October 24, 2001 at 15:55:48, J. Wesley Cleveland wrote:

>There appears to be confusion between the two uses of egtb tables. One is in the
>search where you need to know the value of the position (mate in 37, draw,
>etc.), though for practical play win, draw, or loss is enough. The other is at
>the root where you need to know which move to play.
>
>In the search, Best-move tables would be completely impractical, but a related
>idea might work very well:
>Have an evaluator that predicts the score, win, draw, or loss, and a table that
>indicates correct, or has the real score. With a good evaluator, this table
>should compress *extremely* well. For many unbalanced tablebases, e.g. KQQQk,
>you would not even need the table (i.e. compress it to zero bytes).

I like this one.  If you don't mind not having the distance info you
can get some really impressive compression with this.  Should make it
practical to store all of the 4 piece tables in ram, and if you have a
lot of ram(1-2gig) you might fit all the 5 piece tables in.  You would
still need some other type of table to use at the root since W/L/D tables
don't tell you how to finish the win once you have it, but ...

>At the root, Best-move tables could work well.

Agreed.

Angrim



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.