So far, we have gone through a survey of one-shot games, where players meet just once and pursue the non-cooperative maximization of their respective objectives. In such games, calendar time (either past or future) has no bearing at all on agents’ choices. Moreover, a striking feature of many games is that the resulting Nash equilibria may indeed be Pareto-inefficient (as typically happens in the prisoners’ dilemma). However, casual observation suggests that players – firms, institutions, nations – often do interact repeatedly over long time spans. This prompts the construction of repeated game frameworks in which agents may collude to maximize common objectives, perhaps the sum of their individual ones, though keeping an intrinsically non-cooperative attitude. It is important to stress this latter aspect to distinguish the issue treated in this chapter from what belongs to the theory of cooperative games, which we shall deal with in Chapter 10. Essentially, the difference between repeated games and cooperative games can be grasped in the following terms. While players involved in a cooperative game are assumed to cooperate explicitly towards the maximization of a common objective irrespective of the time horizon (that is, even if the game is a one-shot event), a repeated game relies on the possibility that repeating over time the same constituent game gives rise to a form of implicit collusion based upon non-cooperative rules but nonetheless capable of replicating the outcome of an explicitly cooperative behaviour.