Microscopic Foundation of Stochastic Game Dynamical Equations

Abstract

The game dynamical equations are derived from Boltzmann-like equations for individual pair interactions by assuming a certain kind of imitation behavior, the so-called proportional imitation rule. They can be extended to a stochastic formulation of evolutionary game theory which allows the derivation of approximate and corrected mean value and covariance equations. It is shown that, in the case of phase transitions (i.e. multi-modal probability distributions), the mean value equations do not agree with the game dynamical equations. Therefore, their exact meaning is carefully discussed. Finally, some generalizations of the behavioral model are presented, including effects of expectations, other kinds of interactions, several subpopulations, or memory effects.