Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why do we like it?

Author: David Fotland

Date: 11:19:28 02/02/98

Go up one level in this thread



I like computer go, and now computer chess, for many reasons.  I've been
writing game AI's for 20 years, and expect to continue until I die.

There is a lot of satisfaction in solving a hard problem or puzzle.
Since
game AI's continue to improve, every week or month (or day early in the
life of a chess program :) you get some satisfaction that you have made
an improvement or solved a problem.  There is a rating system, so you
can
see results of your improvements, and can guage your progress.  If you
develop a strong program, you gain fame and recognition among your
peers,
and if you publish a strong program, you gain fame among a wider
audience.
A few people can make a living or significant money from game AI's, but
that's not why most people do it.  In my case, having some income pays
for
new computer hardware every year, and helps convince my wife that it is
ok for me to spend so much time on it.

Chess search techniques (alpha-beta, iterative deepening, hash tables,
endgame databases, etc) are directly applicable to many similar tactical
games, like checkers (see Chinook), othello, or chess variants.  I've
used
them for domineering and othello successfully.  Backgammon is trickier
due to the dice and doubling cube, and the top programs are learning
programs.

In Go, alpha-beta is used for local tactical problems and life and death
analysis.  But since human beginners can look 20 ply or more deep in
some
local situations, go programs use a selective search.  I don't use
iterative
deepening, since I can get very good move ordering with a little go
knowledge,
and the searches have to go very deep in some lines (40 ply or more) to
solve
simple tactics.  Knowledge respresentation and pattern matching are much
more
important in go than in most chess programs.  At the full board, most
moves
are made for strategic reasons, and using an evaluation function and
lookahead
doesn't work well.  Almost all strong programs evaluate moves on
strategic
considerations, rather than evaluate positions with an evaluation
function.

If anyone is interested I can give some examples of positions that
beginning go players solve quickly, that need fairly deep search for
a brute force solver.

David Fotland


On January 31, 1998 at 02:36:31, Stuart Cracraft wrote:

>I've been asked by people unfamiliar with computer chess why
>I enjoy it.
>
>The usual question I give is that it creates "busy work"
>and we all know that the idle mind is the devil's playground. :-)
>In other words, the program seems to have no real ending.
>
>Anyway, why do you like to program computer chess? Do you think
>you could translate some of this into programming another board
>game like Go, Backgammon, Checkers, etc.? Or no?
>
>--Stuart



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.