Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 exptools 1/6/84; site ihlts.UUCP
Path: utzoo!watmath!clyde!burl!mgnetp!ihnp4!ihlts!rjnoe
From: rjnoe@ihlts.UUCP (Roger Noe)
Newsgroups: net.chess
Subject: Re: What makes a strong computer chess program?
Message-ID: <504@ihlts.UUCP>
Date: Wed, 4-Jul-84 11:20:01 EDT
Article-I.D.: ihlts.504
Posted: Wed Jul  4 11:20:01 1984
Date-Received: Thu, 5-Jul-84 00:34:56 EDT
References: <988@eosp1.UUCP>
Organization: AT&T Bell Labs, Naperville, IL
Lines: 20

>	what a really strong chess computer program will be like?
>	. . . chess programs should be able to do something that
>	humans do fairly easily -- analyze long forcing and semi-forcing lines.
>					- Toby Robison (not Robinson!)

That's part of what current chess-playing machines really lack.  Virtually
all such successful machines have quite good tactics but no strategy what-
soever.  They're all just like novice players with prodigious memories and
blinding speed.  What they lack is the learned ability of the grandmaster.
Sure, you can say that it is possible for a machine to play a perfect game
given sufficient speed and storage.  But that's impractical.  Current tech-
niques (basically full-width searches with alpha-beta pruning and evaluation
algorithms which have become rather sophisticated) have pretty much reached
their limit, barring significant advances in hardware technology.  What is
most needed (and I know this is not a new idea in the slightest) is to be
able to program the ability of the expert to select moves that look promising
and reject irrelevant continuations without generating and evaluating posi-
tions several ply down.  Easy to say, hard to do.
--
	Roger Noe			ihnp4!ihlts!rjnoe