On the mean-field limit of diffusive games through the master equation: extreme value analysis

Abstract: We consider an $N$-player game where the players control the drifts of their diffusive states which have no interaction in the noise terms. The aim of each player is to minimize the expected value of her cost, which is a function of the player's state and the empirical measure of the states of all the players. Our aim is to determine the $N \to \infty$ asymptotic behavior of the upper order statistics of the player's states under Nash equilibrium (the Nash states). For this purpose, we consider also a system of interacting diffusions which is constructed by using the Master PDE of the game and approximates the system of the Nash states, and we improve an $L^2$ estimate for the distance between the drifts of the two systems which has been used for establishing Central Limit Theorems and Large Deviations Principles for the Nash states in the past. By differentiating the Master PDE, we obtain that estimate also in $L^{\infty}$, which allows us to control the Radon-Nikodym derivative of a Girsanov transformation that connects the two systems. The latter allows us to reduce the problem to the case of $N$ uncontrolled diffusions with standard mean-field interaction in the drifts, which has been treated in a previous work.