Bridge Bidding via Deep Reinforcement Learning and Belief Monte Carlo SearchOACSTPCDEI

Contract Bridge,a four-player imperfect informa-tion game,comprises two phases:bidding and playing.While computer programs excel at playing,bidding presents a challeng-ing aspect due to the need for information exchange with part-ners and interference with communication of opponents.In this work,we introduce a Bridge bidding agent that combines super-vised learning,deep reinforcement learning via self-play,and a test-time search approach.Our experiments demonstrate that our agent outperforms WBridge5,a highly regarded computer Bridge software that has won multiple world championships,by a per-formance of 0.98 IMPs(international match points)per deal over 10 000 deals,with a much cost-effective approach.The perfor-mance significantly surpasses previous state-of-the-art(0.85 IMPs per deal).Note 0.1 IMPs per deal is a significant improvement in Bridge bidding.