Newsgroups: comp.ai.games,comp.ai,rec.games.bridge
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!newsfeed.rice.edu!bcm.tmc.edu!news.msfc.nasa.gov!newsfeed.internetmci.com!in1.uu.net!news1.digital.com!decwrl!amd!netcomsv!uu4news.netcom.com!netcomsv!uu3news.netcom.com!ix.netcom.com!netcom.com!kwoolsey
From: kwoolsey@netcom.com (Kit Woolsey)
Subject: Re: How can we encourage AI interest in bridge?
Message-ID: <kwoolseyDKFC0I.FuK@netcom.com>
Followup-To: comp.ai.games,comp.ai,rec.games.bridge
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
X-Newsreader: TIN [version 1.2 PL1]
References: <4bursl$79f@pith.uoregon.edu> <4buu08$bud@agate.berkeley.edu> <4bvg9h$fjb@interport.net> <4c15fg$ad3@agate.berkeley.edu> <4c1f4n$e9p@agate.berkeley.edu> <4c1kna$8qn@news.mcn.net>
Date: Sun, 31 Dec 1995 00:00:18 GMT
Lines: 47
Sender: kwoolsey@netcom.netcom.com
Xref: glinda.oz.cs.cmu.edu comp.ai.games:3359 comp.ai:35563

Jim Loy (jimloy@mcn.net) wrote:
: william@beirut.berkeley.edu (Andy Grosso) wrote:
: >Following up on this: I went to the local mathematical
: >bookstore. Books on Game Theory mention backgammon
: >in the preface (as a motivating factor) but then 
: >proceed to ignore it (and other motivating games) 
: >for the rest of the text. 
: >
: >Browsing the web leads to some unreachable sites 
: >and very little information about Backgammon. Certainly
: >almost nothing about "solutions" (or even partial
: >solutions). 
: >
: >Backgammon just doesn't seem like it ought to be that 
: >difficult. And yet, there's this dearth of results. 
: >
: >Is it that no-one's tried to "solve" it (e.g. find
: >optimal moves) ? Or is it that much more difficult than
: >I'm perceiving it to be ? 

: There was an interesting article in Scientific American, many years ago,
: about a program, on a mainframe, which beat a World Champ at backgammon, 
: in a short match.  The author of the article was Hans Berliner, also the
: head programmer on the project, of Carnegie Mellon University.  They've
: got his thesis on file there (or they did).  He was also a former 
: Correspondence Chess World Champ.  

: The program, apparently was world-class, but probably not of world
: championship caliber.  I haven't heard more.  My impression is that 
: backgammon programs on the market are not very skillful, although I 
: haven't bought any lately.


Hans Berliner's program played a decent game, but was definitely not of 
world class caliber.  Its skill level might be comparable to a "bad life 
master" at bridge.

It the last five years, however, there has been a remarkable breakthrough 
in backgammon programs.  Using a neural network approach (where the 
program develops its own strategies by playing out a few million games 
and seeing what is successful), quite strong programs have been 
developed.  I know of at least three such programs which are definitely 
on a par with the best backgammon players in the world, and one of these 
is commercially available.  Will the same approach work with bridge?  
Time will tell.

Kit
