Newsgroups: comp.ai.games,comp.ai,rec.games.bridge
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!newsfeed.internetmci.com!in1.uu.net!ncrgw2.ncr.com!ncrhub6!ncrutr.Utrecht.NCR.COM!news
From: Martin Janssen <Martin.Janssen@ATT.COM>
Subject: Re: How can we encourage AI interest in bridge?
Message-ID: <DLMwGu.KIt@ncrutr.Utrecht.NCR.COM>
Sender: news@ncrutr.Utrecht.NCR.COM
Reply-To: Martin.Janssen@ATT.COM (MARTINJA)
Organization: AT&T WCND Utrecht
X-Newsreader: DiscussIT 2.5.1.3 for MS Windows [AT&T Software Products Division]
References: <4e05fr$aki@calvin.st-and.ac.uk>
Date: Tue, 23 Jan 1996 12:38:53 GMT
Lines: 47
Xref: glinda.oz.cs.cmu.edu comp.ai.games:3576 comp.ai:36228


> ==========A^3, 1/22/96==========
> 
> Martin Janssen (Martin.Janssen@ATT.COM) wrote:
> : The real challenge is to define a good strategy for a bridge
> : playing program. After all, what is a good strategy. Start by
> : taking as many quick tricks as possible by playing Aces, Kings,
> : etc., is not the right strategy, as beginners very soon find
> : out. On the other hand, the purpose of the game is to make as
> : many tricks as possible. So how do you program this? How do you
> : evaluate the play so far at any given point in the play? In
> : other words: what is a good strategy for playing bridge, that
> : works in all cases?
> 
> An evaluation strategy would, I think have to be based on
> probabilities, and tak einto account the current state-of-play
> in the rubber as well as the cards previously played in this
> hand and the bidding of the hand. There are two distinct
> types of play - playing to make a contract, and playing
> to defeat a contract. One may make a risky play to get two 
> over-tricks at best or make the bid at worst, but one shouldn't
> make a risky play to make on over-trick at best and one under
> at worst, if there is a strategy that will ensure that the
> bid is made exactly. Again, if you have a chance at making 
> the bid at best or going two down at worst, or can definitely
> make one under with a different strategy, the usual play is to
> take the risk - there's little difference between one and two under,
> but a world of difference between one under and making the bid.
> Similar situations apply in reverse when trying to defeat a
> bid. This sort of complexity is of course what make Bridge on of the 
> most interesting and difficult whist games.

You're right. The tactics are different for duplicate bridge,
robber bridge etc. Also the "strategy" for playing in a given
suit is different from playing a NT contract. So a good program
would have to take all this in consideration. But even without
this complication it would be far more complicated IMHO,  to
write a program that plays bridge reasonably well, compared to
e.g. a chess program. 
> 
> --
> TTFN, A^3 ***************E-mail*aaa@dcs.st-and.ac.uk*****************
> ***Mundus Vult Decipi****S-mail*40 Fife Park, St Andrews KY16 9UE****
> ****************************Tel*+44-1334-463268***+44-589-464141*****
> ********Home Page: <http://www-theory.cs.st-and.ac.uk/~aaa/>*********


