Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Want tips and tricks on tablebase CONSTRUCTION

Author: Simon Finn

Date: 11:47:32 10/08/01

Go up one level in this thread


On October 08, 2001 at 10:27:34, Guido wrote:

>Reading the Nalimov message you can understand why it is necessary compute two
>endings at the same time, i.e. with move to White and move to Black.

If you do the two endings simultaneously, you have to look only
one ply forward to compute the new minimaxed value for each position.

>If you compute one ending at a time, you multiply the cpu time by 10-20.

If you do one ending at a time, you apparently have to look forward
two plies to compute the new minimaxed value. The factor of 10-20
is (roughly) the branching factor due to the extra ply.

Actually, there's a trick that works for DTC databases (not sure
about DTM databases) that involves looking one ply forward and
one ply backward (from positions with the "wrong" player to move)
that reduces the CPU penalty to a factor of 2. In the right
circumstances, this could be a good time/space trade-off.

Simon



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.