The 1997 world champion of computer bridge uses AI planning techniques to plan its declarer play.


Bridge Baron is a computer program that plays bridge. It won the 1997 world championship of computer bridge, the Baron Barclay World Bridge Computer Challenge, as reported in The New York Times and The Washington Post. The five-day competition, which was hosted by the American Contract Bridge League in July 1997, included five computer programs, from the US, Japan, and Germany. The Bridge Baron won every head-to-head match that it played against the other programs.

Copies of Bridge Baron are available from Great Game Products of Bethesda, MD.

AI Planning in Bridge Baron

We developed an AI planning algorithm for generating and evaluating strategies for declarer play in the game of bridge. As shown in the figure, the algorithm generates alternative strategies for how to play the game, and evaluates these strategies to determine which of them is best.

We incorporated our implementation of this algorithm into Bridge Baron, an existing commercial product owned by Great Game Products. This helped Bridge Baron win the 1997 championship, and it has been used in all subsequent commercial releases of Bridge Baron.

The planning algorithm is a special case of a technique called ordered task decomposition that can be used in a wide variety of planning problems. For example, the SHOP2 planning system, a domain-independent planner using ordered task decomposition, won a prize at the AIPS-2002 International Planning Competition.


By us:

By others:


The AI planning techniques used in Bridge Baron were developed by Stephen Smith, Dana Nau, and Tom Throop. Here is how to contact us:

Stephen J. J. Smith and Tom Throop
Great Game Products, Inc.
7825 Tuckerman Lane, Suite 206
Potomac, MD 20854

Tel: 800-426-3748 or 301-299-9005
Fax: 301-765-8004


Dana S. Nau
Department of Computer Science
and Institute for Systems Research
University of Maryland
College Park, MD 20742

Tel: (301) 405-2684