Subgame perfection in Positive Recursive Games with perfect information

We consider a class of n-player stochastic games with the following properties: (1) in every state, the transitions are controlled by one player; (2) the payoffs are equal to zero in every nonabsorbing state; (3) the payoffs are nonnegative in every absorbing state. We propose a new iterative method to analyze these games. With respect to the expected average reward, we prove the existence of a subgame-perfect epsilon-equilibrium in pure strategies for every epsilon > 0. Moreover, if all transitions are deterministic, we obtain a subgame-perfect 0-equilibrium in pure strategies.