Author: Guido
Date: 05:22:48 03/26/03
Go up one level in this thread
On March 26, 2003 at 04:29:07, martin fierz wrote: >On March 25, 2003 at 19:29:27, Dieter Buerssner wrote: > >>On March 25, 2003 at 17:30:52, martin fierz wrote: >> >>Not exactly an answer to you question. But when you miss just some percents of >>memory, you could try to use 1.6 bit per position. 3^5 < 2^8, so it is easy to >>store 5 win/draw/loss results in one byte. I think, the additional overhead of >>extracting the "trits" (0/1/2) is not that serious. >> >>Regards, >>Dieter > >2 bits per position are used to represent not only win/loss/draw but also >unknown during the generation process. so you have to encode 4 values instead of >3 :-( It is not necessary because you can use draw also as unknown during generation. At the end the draw-unknown assume the meaning of a draw. In this way clearly you have not a value for illegal positions. Ciao Guido >naturally, the final database only needs to store win/loss/draw, and there of >course i use something like what you suggest. > >cheers > martin
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.