We propose a solution concept called Generalized Sampling Equilibrium (GSE), where players use statistical rather than strategic reasoning. This concept is rooted in the sampling equilibrium of Osborne and Rubinstein (1998, 2003), and accommodates a variety of other statistical inference procedures. We show that the GSE is unique for a large class of two-action games, and characterize how it relates to the Nash equilibrium. We also characterize how the GSE changes with the size of the sample players obtain, and demonstrate the predictions of this solution concept in several applications including a labor matching environment. We show that sampling introduces a friction that results in larger unemployment than in Nash equilibrium.