The popular card game DouDizhu in China has become a research hotspot of computer game because of its unique characteristics. For the bidding stage of DouDizhu game, accurate evaluation of hand cards strength is the key to mastering accurate bidding. In this paper, we propose a method combining convolutional neural network and reinforcement learning, and add the feature of winning distance. Supervised learning is used to train agent, and perfect information distillation technology is used to optimize bidding methods.The experimental results show that the bidding network model proposed in this paper is effective, and by comparing with some open source DouDizhu AI, it proves that the bidding method in this paper is effective.