Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: EGTB question to Martin Fierz

Author: Dieter Buerssner

Date: 10:21:08 04/19/02

Go up one level in this thread


On April 19, 2002 at 12:49:10, Alvaro Jose Povoa Cardoso wrote:

>I'vv been studing Jonathan Schaeffer's EGTB paper and I got stuck in the
>indexing functions.
>Before I red the paper I had a basic vague idea on EGTB indexing.
>My naive scheme was reserving 5 bits for each piece. So for all 4 piece
>databases I have 20 bits. This scheme has indices larger than the number of
>positions possible. But what about 8 piece databases? I would need 40 bits.
>This just don't look good.

I am not Martin, and I neither do no much about checkers. But you seem to miss
one idea: that the pieces are indistinguishable. Arguing for Chess and an
illegal QQ position (both Qs same color, no Ks for this example). How many
different positions are there? Ignoring symmetry, the naive answer would be
64*64 (or 64*63, doesn't matter for this). But Qa1,b1 is the same as Qb1,a1. So
you only need 64*63/2 indices. For 3 Qs, it would be 64*63*62/3!, etc. You might
ask, how to index them then. For example by a precalculated helper array:

index = precalculated_index[sq1][sq2]... // instead of sq1 + sq2*64 + ...

Regards,
Dieter



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.