Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Arasan 5b: congratulations to John

Author: Jon Dart

Date: 14:40:36 03/24/99

Go up one level in this thread



Certainly 5.0b is better than 4.1, which was the previous released
version. Whether it is master level or not depends perhaps on the
hardware. On a 233 MHz machine, which I use on FICS, it is still
losing games to human masters and experts, especially at slower
time controls. It wins some, too, but it is far from invincible.

I tend to think of a "knowledge-based" program as one that uses chess
knowledge to guide the search, either by selectively pruning out
"uninteresting" moves (Shannon Type B strategy) or extending lines
based on something more sophisticated that the standard check and
capture extensions. In this sense, Arasan is not a knowledge-based
program. It has a fairly complex evaluation function and "understands"
many positional concepts, but it uses this knowledge inside a fairly
standard search framework with commonly-used extensions.

I have a newer version with a much faster hashtable implementation
and a couple of bug fixes, which I am hoping to release in the next
month or so. It has been doing well on my test suites and in games.

--Jon

On March 24, 1999 at 10:32:38, Fernando Villegas wrote:

>Hi John:
>I have been following your sucesive Arasan incarnations and I can say you that
>it seems you have reeached the level where Arasan can be considered at least as
>fide master and maybe more. I have played one game against it at 40 moves in 90
>minutes and I saw great, great progress. Tha game finished in a draw and it was
>great fun to play it.
>Now some questions: it is yours program, as it seemed to be to me, a knowledege
>based one?
>I look forward for your answer and even better versions!
>fernando



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.