Fig 4. Heat maps of relative cooperation for Scoring, Shunning, Simple Standing, Staying, and Stern Judging, i.e. each cell represents the average amount of cooperation under Bayesian reasoning minus the average amount of cooperation under non-reasoning for different error rates.

The average is over initial conditions evenly spread across the simplex and r = 3. Note that reasoning generally under-performs compared to non-reasoning except for Scoring.