Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Help on rewrite decision (ajedres tactico)

Author: Dann Corbit

Date: 21:09:25 06/18/04

Go up one level in this thread


On June 18, 2004 at 23:55:53, Cesar Contreras wrote:

>Hi
>
>I'm stuck with my program "ajedrez tactico", so i'm planing a complete rewrite,
>i have the dream of creating a top class chess engine, i think i can. I have
>many ideas in my mind,  but a fundamental decision i have to make it's to chose
>the board representation: Bitboards or 0x88 or anyother?

Those two are the best.  Actually, many top programs use both 0x88 and bitboard.
You can be successful with either.

>Currently i use a 64 bytes array with a Mailbox move scheme, that's why i think
>i could be more confortable with 0x88, but i don't know if this it's good
>enought to make a top class chess program. Are there many (or any) class
>programs that currently use 0x88?

0x88 programs have been #1 on the SSDF list.

>Maybe it's an already answered question, but i'm new to this forum, so if you
>don't want to waste much time, just post your choice (like a poll): Bitboards or
>0x88?

Do the one you enjoy the most.  The strength of your chess engine will be made
by the algorithms you choose.  The board representation may make for 15%
difference in speed one way or another (I think that 64 bit CPUs make bitboard
look better than before).  But it might be possible to write 0x88 that is as
fast or faster than bitboard on 64 bit machines.

Far more important is your choice of algorithms.  A bad choice in algorithms
will make your program literally millions of times slower.  For instance, if you
use pure mini-max, you won't make it past 7 plies on a very fast machine.  Just
a good alpha-beta implementation will drastically improve your search depth (and
strength of play).

Having code that is carefully debugged and clear will also be very helpful.

Start small and build up.  A very small program can be very powerful.  Look at
Olithink (open source) and Thinker (tiny executable is about as strong as the
top amateur programs).  You don't need 10,000 features right away.



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.