We introduce a two-player model of reinforcement learning with memory. Past actions of an iterated game are stored in a memory and used to determine player’s next action. To examine the behaviour of the model some approximate methods are used and confronted against numerical simulations and...