Current studies mainly focus on the cooperation and competition between decision making unites(DMUs) and evaluators but neglect the non-cooperation and competition among DMUs in the process of the evaluation. With regard to the gap and combined with the theory of cross competition game, the evaluation issue in this paper is defined as the cooperation and competition between evaluator and DMU, and the cross competition game between two DMUs. Considering the characteristics of DMUs indexes which are not fixed but adjusted dynamically, the rules of the cross competition game are designed, and the data envelopment analysis(DEA) evaluation and selection process of the cross competition game are described by using the decision tree method. Simultaneously, we change the route of utility values in the evaluation process from only considering weights exchange to considering cross competition game. Furthermore, an improved cross efficiency DEA model and the classification of deterministic, risky and uncertain DEA are proposed, and an approach of game efficiency DEA evaluation is designed. For the proposed model and approach, we explain them from the perspectives of the game theory in economics and decision making theory in management. Finally, an illustrated example is provided to demonstrate the feasibility and effectiveness of the proposed model and approach.