Premium
Declarer: A learning bridge‐playing program which generalizes and infers
Author(s) -
Napjus Chris N.
Publication year - 1971
Publication title -
behavioral science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.371
H-Index - 45
eISSN - 1099-1743
pISSN - 0005-7940
DOI - 10.1002/bs.3830160405
Subject(s) - bridge (graph theory) , bidding , computer science , process (computing) , subroutine , line (geometry) , adversary , artificial intelligence , sequence (biology) , function (biology) , basis (linear algebra) , machine learning , programming language , mathematics , computer security , medicine , geometry , marketing , evolutionary biology , biology , business , genetics
A bridge program has been written which learns to play any combination of declarer/dummy cards in a single suit, given initially only the basic rudiments of the game. The learning process emulates the manner in which a person learns to play bridge, building upon increasing experience to generalize upon familiar situations, to recognize significant card patterns, and to alter a planned line of play on the basis of new information gained from an opponent's play. Opponents' plays are generated automatically by a subroutine. Both the overall line of play and individual trick play decisions may be modified on the basis' of inferences which are drawn concerning the location of cards in the opponents' hands. Estimates of these missing cards, for all four suits, are made initially by analyzing each bidding sequence, and are updated as each card played either reinforces or negates confidence in the current estimate. A confidence score is maintained for each estimate as a function of the amount of information available on which to base it. Deviations from the planned line of play occur only when the confidence in the current estimate exceeds the experience‐determined likelihood of success of the planned play.