Sophisticated imitation in cyclic games

Consider a large population of individuals that are repeatedly randomly matched to play a cyclic 2 x 2 game such as Matching Pennies with fixed roles assigned in the game. Some learn by sampling previous play of a finite number of other individuals in the same role. We analyze population dynamics under optimal boundedly rational behavior (in the sense of Schlag, 1998c). We find that long run play is close to the Nash equilibrium (when few individuals receive information) if and only if the sample size is greater than one. (orig.)Available from TIB Hannover: RO 3009(427) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekSIGLEDEGerman