In an experiment on the repeated prisoner’s dilemma where intended actions are implemented with noise, Fudenberg et al. (2012) observe that non-equilibrium strategies of the "tit-for-tat" family are largely adopted. Furthermore, they do not find support for risk dominance of TFT as a determinant of cooperation. This comment introduces the "Payback" strategy, which is similar to TFT but is sustainable in equilibrium. Using the data from the original article, we show that Payback captures most of the empirical support previously attributed to TFT, and that the risk dominance criterion based on Payback can explain the observed cooperation patterns.