bart@dogmatix.cs.uoregon.edu (Bart Massey) (04/18/91)
Here's the promised summary of responses to my request for bibliography/info on AI/probabilistics in bridge play. Thanks much to the following people for their helpful and interesting information: Sam Dinkin <sdinkin@tybalt.caltech.edu> <schaffer@cs.rutgers.edu> Steve Maurich <maurich@tigger.colorado.edu> Frode Fjeldbo <frodef@idt.unit.no> Micah Fogel <fogel@math.berkeley.edu> David Dowe <dld@stan.xx.swin.oz.au> Bjarni Hjardar <bjarni@be.chalmers.se> Alex Martelli <alex@am.sublink.org> Devdas Nandan <nandand@clutx.clarkson.edu> I'll try to follow up some more of the leads I was given, and do some more library research, and if I find out anything worthwhile, I'll definitely let you know. Bibliographic Info: Articles: (Note that I didn't find much in the standard indices in the library, but that may just be my clumsiness as a researcher... :-) Learning From Textbook Knowledge: A Case Study William W. Cohen AAAI '90 Proceedings vol. 2 pp. 743-748 ISBN 0-262-51057-X This article describes a system which learned how to make good opening bids using conventions hand-translated from Sheinwold's "5 Weeks To Winning Bridge" and augmented by carefully selected examples. A very nice article, but more of an AI study than a practical approach to bridge bidding. Graham Menhennit M.Sc. Thesis University of Melbourne Parkville, Victoria 3052, Australia approx. 1979 I haven't yet obtained this -- if someone in OZ could mail me a copy, I'd gladly pay postage, and be very grateful. Anonymous Ph.D. Thesis "AI Bridge Declarer System" University of Washington Seattle, WA approx. 1980 If someone at UW could find this for me, I'd be very grateful. Otherwise, I'll try to chase it myself RSN. Books: Turbo Pascal Game Works 1985 Borland International Inc. ISBN 0-87524-146-8 Contains (among other things) a description of the inner workings of their Pascal program "Turbo Bridge", whose source is also available from Borland as part of the above-mentioned "Turbo Pascal Game Works" package. This program uses hard-coded bidding rules which work only moderately well. The play is via a Monte Carlo approach, which is quite fast but prone to error -- IMHO it really only makes sense on an 8088, where probabilistic calculations are way too expensive. The Mathematical Theory of Bridge Emile Borel and Andre Cheron tr. ed. Alec Traub Apparently THE reference on bridge mathematics, but kind of hard to find. One respondent gave me some useful leads, but I haven't chased them down yet -- e-mail me for details if you are searching for this. Bridge Odds for Practical Players Hugh Kelsey and Michael Glauert Master Bridge Series Victor Gollantz Ltd., London, 1980 ISBN 0-575-02799-1 This is probably a more useful book for human players than Borel's, but less useful for computer programmers. Comments: Everyone seems to agree that surprisingly little progress has been made on the play problem specifically -- most of the published work is on bidding, which is a different sort of problem. A number of people are also working on various aspects of bridge play. I probably shouldn't publish their names without asking them, but e-mail me if you're interested in getting in touch with these folks. Again, thanks much to all who responded, Bart Massey Dept. of Computer Science University of Oregon Eugene, OR 97403 (503) 346-4156 bart@cs.uoregon.edu