A non-equilibrium analysis of the finitely-repeated prisoner's dilemma