Anonymous repeated games with a large number of players and random outcomes (Q750328)

The paper considers anonymous repeated games \(H^{\infty}\) with the one shot game H defined by \((K,\mu,\bar A,X,y,F)\) where, (1) \(\mu\) is a probability measure on the set K of players, (2) \(\bar A\) refers to the set of actions available to each player, (3) X refers to the set of possible actions (signals) of the game, (4) \(y=(\pi_ k:\) \(\bar A\times X\to R)_{k\in K}\) describes the payoff functions of the players, (5) F is a function describing the way of arising the signals, which is dependent on the players' strategies \(a\in A=\bar A^ K\). This is defined by \(F(a)=G(m(a))\), \(a\in A\), where m(a) are the measures of the form, \(m(a)(B)=\mu \{k| a_ k\in B\}\) for \(B\subset \bar A\) (here \(a_ k\) is the i-th component of a), and G is a function transforming measures m(a), \(a\in A\), into the set M(X) of probability measures on X. The total payoff in \(H^{\infty}\) is the discounted sum of the one shot payoffs with a discount factor \(\delta\in (0,1)\). The author shows under additional assumptions that any subgame perfect equilibrium in \(H^{\infty}\) determines an \(\epsilon\)-Nash equilibrium in H. Besides, the case when the convergence of a sequence of games \(\{H_ n^{\infty}\}\) to its continuous representation \(H^{\infty}\) implies the convergence of some vectors of actions from \(H_ n^{\infty}\) to a Nash equilibrium in \(H^{\infty}\) is characterized.