Due to the enhancement of web service and people's love needs, social networking service is emerging and getting more and more popular in recent years. Recently, social games have been drawing much attention. Social games contributed more people to play online games, and to extend their social networks. The popularity of social games also changed the operation model of online games. High production cost online games are threatened by low-cost website-recommended-only social games. Thus, it is very important for the on-line game service providers to find out the key factors influencing player's decision. Currently, there are various kinds of social games, for instance, harvest and farm (Happy Harvest, Country Story, and Farm Vills), Pet breed (Pet Society and My Fishbowl), and the category of Restaurant and inn management (Restaurant City and Café World). In this study, the Fuzzy cognitive map (FCM) is used to construct the network relation map (NRM) and the extended NRM, the Principal component Analysis (PCA) is applied to categorize criteria based on their properties, the Analytic network process (ANP) is used to determine the weights of components, and the VIKOR (Vlse kriterijumska Optimizacija I Kompromisno Resenje) is adopted to evaluate the performance gaps of social games. Service provider of social games could clarify the relationship between the service performance and the behavior of use, and obtain improvement strategies of service performance for social game service.