This study employs the two-production-stage data envelopment analysis (DEA) proposed by Seiford and Zhu (1999), in order to measure the efficiency of nineteen online game companies in Taiwan and mainland China during 2009-2012. The approach for adjusting negative inputs or outputs proposed by Ali and Seiford (1990) is adopted. The empirical result shows that online game companies in Taiwan show a downward trend in the first stage (production stage); however, online game companies in mainland China show an upward trend in production efficiency. Online game companies on both sides have upward trend in both the second (profits stage) and aggregation stage. Besides, the Tobit regression model is used to estimate the effects of environmental variables on efficiency. The result shows that founding years positively influence efficiency in both the first stage and aggregation stage. The R&D intensity positively influences efficiency in among the first stage, the second stage and aggregation stage. The liability ratio negatively influences efficiency in first stage. Compared to those in Taiwan, online game companies in mainland China’s benefit from locational advantage. Marketing intensity and firm size negatively influence efficiency in both first stage and aggregation stage. At last, this study employs BCG matrix for providing useful decision-making information.