Fig. 1.

(A) A round of the steal/punish game (5). $s$ is the value of the stolen good, $c$ the cost of punishing, and $p$ the cost of being punished. $s,c>0$ and $p$ > $s$. The game is repeated for $N$ rounds. (B) Pure strategies in the steal/punish game. Typical strategy spaces in models of retaliatory punishment, depicted by the yellow area, produce the familiar equilibrium of rigid punishment and flexible theft. Extending this strategy space to allow flexible victims reveals an inverted equilibrium: rigid theft and flexible punishment. The direction of selection hinges on the outcome of flexible thief against flexible victim: Whichever role tends to “back down” first evolves rigidity to compensate. (See Fig. S1 for payoffs.)