Academic journal evaluation is one of the important issues in journal research field. Traditional methods of data envelopment analysis are based on self-evaluation when evaluating the performance of academic journals. Without peer-evaluation,the evaluation result might be easily exaggerated. This paper studies the problem under the non-cooperative game framework by using data envelopment analysis /assurance region ( DEA/AR) game cross efficiency method,and 17 Chinese sci-tech core journals are selected as an illustrative example. The results show that: 1) Some academic journals with high industry influence have non-ideal evaluation results: although their outputs are ideal,their published papers are non-ideal. 2) The evaluation values of academic journals could be increasing,unchanged,decreasing with the change of the preference of index weights. 3) The self-citeing of journals have three cases: excessive,moderate and insufficient. The evaluation result can provide a powerful decision reference for journal assistants to improve the competitiveness of journals.