We study a class of stochastic linear-quadratic differential games played between known and unknown rivals. For two known rivals, we find all time-dependent Markov perfect Nash equilibria that are linear in the state. Then we consider unknown rivals, by assuming that players can only condition on the state and are unaware of the other players in the game. With unknown rivals, we show that if there are multiple equilibria players almost surely learn an inefficient equilibrium.
Bienvenue à tous!