In order to solve the problems of long competition system and hidden danger of cheating in the current competitive two-against-one competition, a system of card strength evaluation and equal card strength generation is designed. Carry out clustering analysis on a large amount of human card play data by using a second-order clustering algorithm to obtain a labeled data set of hand card strength classification; A long short term memory network (LSTM) model based on attention mechanism is constructed to train the hand sequence and hand strength, and the hand with the same hand strength is generated to conduct a sparring experiment with the help of AI robot. The experimental results show that there is a significant difference in the winning rate between the generated different card strength hands, and the corresponding card strength hands can be generated correctly, which is feasible.