Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: EGTB algoritmic question

Author: Alvaro Jose Povoa Cardoso

Date: 10:56:46 07/24/02

Go up one level in this thread


On July 24, 2002 at 13:48:45, Dieter Buerssner wrote:

>On July 24, 2002 at 13:34:54, Alvaro Jose Povoa Cardoso wrote:
>
>>Hi all,
>>I've been working on EGTB generation with DTM.
>>I use the 2 most significant bits of a byte to store win/draw/loss/broken.
>>So I have 6 bits left for DTM that gives only 64 values wich are insufficient
>>for my needs.
>>My question is if there is a way to increase this 64 DTM limit and still use a
>>single byte for each EGTB entry.
>
>Sure, there is. By using one high bit for broken and loss, you lose lots of
>space in the low bits. So and obvious scheme, that has DTM/draw/broken, could
>use just 2 values for loss and broken. This leaves 254 values in a byte for the
>mates. These values could be mated/mated in 1/mated in 2/.../mated in 126/mate
>in 1/mate in 2/mate in 127. Actually, I think this (or very similar) is what
>Nalimov TBs do. One could easily get rid of broken and mated (in 0), when one
>has a trusted caller, which would give just one distance to mate more
>
>Regards,
>Dieter


Could you please give me an example: if at a node we have 2 moves such that one
is a loss in 100 and the other is a win in 100. Could you please tell me those 2
byte values?

Thanks,
Alvaro




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.