A characterization of sequential equilibrium strategies in infinitely repeated incomplete information games (Q1124544)

The paper investigates properties of sequential equilibria strategies in infinitely repeated standard two-sided incomplete information games, with the following structure: There are \(| K| \times | R|\) pairs of matrices \(\{(A^{kr},B^{kr})| k\in K\), \(r\in R\}\), which are the payoff matrices of Player 1 and 2, respectively. A pair (k,r)\(\in K\times R\) is chosen according to a distribution on the set \(K\times R\). Player 1 is informed of k and Player 2 of r, and then the bimatrix game \((A^{kr},B^{kr})\) is played infinitely many times, provided that at each moment the history of the players' choices becomes their common knowledge. The players' payoffs in the whole game are defined as the discounted sum of their winnings at each step. A simpler interpretation of the equilibrium behavior is found and it is shown that the set of sequential equilibria coincides with a family of stationary chains determined by a triplet of functions (x,y,T), which are defined on a state space of player type distributions and payoffs (x and y describe a way of behavior of Player 1 and 2 at each stage, respectively, T is a stationary transition function). The evaluation of any play of the game implied by any sequential equilibrium is completely determined by that triplet and an initial state.