This research introduces a new approach to evaluate thermal power plants. In this study, we define two categories of inputs (operational and non-operational) to measure performance of power plants. In a real case study, it is necessary to explore how to combine two separated efficiency measures in a unified structure. The conventional data envelopment analysis (DEA) may fail to discriminate among decision making units (DMUs) and the resulting efficiency scores may not be meaningful, especially when the number of DMUs is insufficient. In this paper, a new approach is presented based on DEA and game theory to evaluate DMUs by a large scale of measures. For this purpose, bargaining game as a cooperative game model and the conventional DEA models are combined. The proposed approach regardless of the number of DMUs does discriminate among the DMUs more effectively. Moreover, DMUs do be compared by different categories of measures in the competitive environment. Furthermore, the case study of thermal power plants in Iran is presented to show the abilities of the proposed approach. Power plants usually have the governmental structure in Iran. Finally, we answer the question of "why the power plants should move towards becoming private in Iran?". (C) 2011 Elsevier Ltd. All rights reserved.