A Probability of One-Shot Cooperation for All Parameters and Treatments

The gures over the next few pages show the results from all runs of the simulations as described in the main text. 1 Figure A.1: Repentant and forgiving strategies decrease one-shot cooperation when one-shot games are very rare (P = 0:1) and agents have 10 partners. These show the expected probability of one-shot cooperation averaged over the last 500 generations of the 10,000 generation simulation for all values of d, b, and w. These are the same parameter combinations reported by Delton et al. (2011). Treatment 1, where agents, as in DKCT play only TFT or ALLD has the highest probability of one-shot cooperation. Treatment 2, where ALLD is replaced by a repentant strategy, DIMAS, has less one-shot cooperation. In Treatment 3, where a savvy repentant strategy, HGRIM, competes with TFT and TF2T, one-shot cooperation virtually disappears. And Treatment 4, where HGRIM is replaced by HTFT. This highlights the variability in outcomes for the same environmental parameters. 2 Figure A.2: Repentant and forgiving strategies decrease one-shot cooperation whe