The tactical-level wargames system has strict structure and strong confrontation, which can be used as a deductive platform for studying military intelligent decision-making. In this paper, an intelligent action decision-making framework of tactical-level wargames is designed, and under the guidance of this framework an wargames intelligent engine is desinged. The engine utilizes a data mining method to extract evaluation attributes such as the historical position probability, the control degree and observation degree from the historical data of the wargames competition, and uses the multi-attribute comprehensive evaluation soft optimization algorithm and the wargames’ basic rules to decide actions. The experimental results show that the engine has the advantages of strong real-time performance, high flexibility and good portability, and has achieved excellent results in the machine-machine and human-machine challenge.