Study: Random reward fluctuations reshape outcomes of classic game theory contests
A new mathematical model shows that even small random variation in payoffs can dramatically change the stable outcomes of classic games like the prisoner's dilemma and rock-paper-scissors. The findings were published in Physical Review Letters.
Researchers built a mathematical model to examine how classic game-theory contests behave when the rewards for different choices vary randomly over time, rather than staying fixed as in traditional versions of these games.
Classic games revisited
In the standard prisoner's dilemma, where two captured suspects must decide whether to cooperate or betray each other, the model normally settles on a single stable outcome: everyone betrays everyone. But when the researchers introduced even modest random variation into the rewards over time, a second stable point emerged, allowing cooperators and defectors to coexist. Under greater variation, the all-defector outcome became unstable, leaving only cooperation as the stable strategy.
Similar effects appeared in the game of chicken. Without variation, the stable outcome is universal swerving — survival for all. Adding a bit of randomness produced a population that doesn't swerve, and further variation led to an unstable flipping between survival and collision.
In rock-paper-scissors, which normally has no stable point at all — strategies simply keep cycling among the three choices — random reward variation introduced new stable and unstable points. Depending on the setup, this could speed convergence toward continuous cycling or, when payoffs are uneven between choices, produce a predictable, repeating limit cycle.
Implications
The researchers conclude that a changing environment can dramatically affect the optimal strategy in a game, even when players' underlying behavioral tendencies stay the same. This may help explain why cooperation persists in real life even in situations that classic theory predicts should favor betrayal, since external forces constantly reshape the actual rewards at stake. The study was published in Physical Review Letters.
