Figure 5: Evolutionary dynamics in the space of stochastic memory-one strategies,.

(a,b) show the marginal distribution of the evolving cooperation probabilities p_ij in the mutation-selection equilibrium. To generate the figure, we have calculated the invariant distribution for a discretised version of the state space, using a grid size of δ = 0.2. For low costs, the cooperation probabilities are in line with WSLS behaviour; for larger cost values, cooperation breaks down, and most evolving strategies are self-defectors. In (c) we depict the strategy that has the highest linear coefficient L(p) according to Eq. (13). Again there are three parameter regions: for low costs, a variant of WSLS is most favoured by selection; for intermediate costs, the somewhat paradoxical strategy (0, 1, 0, 0) is most favoured; and for high costs, AllD becomes predominant. Parameters are the same as before: Population size N = 100, ε = 0.01, and w = 10.