Game-dynamical aspects of the prisoner's dilemma (Q1121186)

This paper is devoted to the study of the iterated prisoner's dilemma with a random number of rounds of the game. The probability that the game is extended to another round is a constant \(w\in [0,1]\). In every round both players have two pure strategies: to cooperate or to defect, while in the game process some stochastic strategies are used. Next the expected payoff function is constructed in single round and in the full game. For the future it is assumed that only a finite number of stochastic strategies, say n, is used by the players and the corresponding payoff matrix is known. Then the game dynamics is described by a system of n autonomous, nonlinear differential equations on the unit simplex \(S_ n\) as the (invariant) state space. For small n (e.g. \(n=2,3,4)\) the orbits of the above dynamical system are studied. Some results of numerical experiments together with the pictures (for \(n=3)\) are presented.