Can a single defector destroy a world of cooperators? The lone red defector at the center grows into a symmetric pattern of blue, red, green and yellow that keeps changing after it wraps around the torus.
Try this Set temptation to 1.8 and press Reset (temptation is read every tick, but the single defector only starts over on Reset). The defectors spread for about 30 ticks and then freeze into a fixed symmetric pattern that leaves cooperators 86% of the board, far more than at 1.65.
About this model Where it ends up
Nowak and May's spatial prisoner's dilemma (Nature, 1992), in the form used in the NetLogo Models Library. Players on a grid have no memory and no strategy beyond cooperate or defect. Each plays its neighbors and then copies whichever strategy scored best nearby, and cooperators can survive in clusters where in a well-mixed population they would die out.
Each patch of a 61 × 61 torus is a cooperator or a defector, and each tick has two stages, each applied to all patches at once. First every patch scores its games against its 8 neighbors (not itself): a cooperator gets 1 per cooperating neighbor and nothing from defectors; a defector gets temptation per cooperating neighbor and 0.01 per defecting one. Then every patch takes the strategy with the highest score among itself and its neighbors, ties going to cooperation. Colors follow Nowak and May: blue stayed a cooperator, red stayed a defector, green switched to cooperating, yellow switched to defecting. seed-mode and initial-cooperation are read at setup, temptation every tick. The single-defector start involves no randomness, so every run with the same temptation is identical.
Nowak and May also counted each cell's game against itself. This port doesn't, so the temptation values that separate its regimes differ from the ones in their paper: here 1.6 to 1.65 gives the shifting pattern, while 1.75 to 1.9 freezes within about 30 ticks.
