campbell@aerospace.aero.org (L. Andrew Campbell) (05/25/89)
In article <134@sierra.stanford.edu> >I am involved in a project to develop an expert system to bid. Since >there has been considerable discussion about PC bridge-playing >programs of late, I thought it appropriate to post this request. I >was wondering if any readers had been involved or seen the results of >previous attempts to apply artificial intelligence to bridge bidding. >I would be particularly interested in previous attempts to develop >expert systems and how these systems fared with particular types of >hands. I am not at all interested in the programming details, only >how existing AI systems bid. > >Feel free to e-mail. If you do, note whether or not your answer can >be summarized for a later posting (I shall do so if interest >warrants). > This may be a bit far afield, as it concerns play, not bidding. I have a technical report from the Weizmann Institute of Sciences, Dept. of Applied Mathematics, by Yossi Nygate, titled "Python: A Bridge Expert on Squeezes" (CS84-27, December 1984). The system is coded in Prolog. Not all the user interface details are given, but a lot of Bridge related code is there, mostly involving Love's squeeze conditions (BLUE). A generate and test paradigm is at the heart of the program. Some planning is involved. I am a "fan" of both Bridge and Prolog (AI language - for other net readers), and would like to see some discussions of the basic algorithms one could use for implementing parts of a bridge system: e.g. double dummy problems, suit combinations (see the Bridge Encyclopedia), strategies (crossruffs, endplays), bidding conventions, hand evaluation, etc. Either pseudo-code or simple prolog (perhaps with some unimplemented predicates) would be best for describing these in some ultimately implementable way. As a challenge, how would you program a brute force solution to the Whitfield Six problem? (See any good history of bridge double dummy problems). I am replying via news (rather than Email) to stimulate further discussion. I realize that the above is not truly responsive to the original posting. Followups to rec.bridge only (Not to comp.lang.prolog). In any case, let us hear more about the Bidding Expert as it evolves.