The purpose of this research is to identify the determinants of loyalty in online games. This study developed a research model to analyze the factors explaining the loyalty level from gamers, employing social identification, flow, and positive anticipated emotion as major research variables, and collected 1308 survey responses from gamers. Within the context of arguing that the exploration of age range issues with respect to online games is important, this research also examines the age differences in path coefficients. To this end, the structural model was tested with the data from entire data sample (i.e., the age of 10s, 20s, and 30s pooled together) and each of the subsamples (i.e., teens taken separately, twenties taken separately, and thirties taken separately). Properties of the causal paths, including standardized path coefficients, the significance of difference, in the hypothesized model, are also presented, so that we can investigate the relative influences of different dominants, demonstrating how teens, twenties, and thirties differ in their decision-making processes regarding the flow, social identification and loyalty from online games.