Why the second-best strategy wins the tournament

In repeated-game tournaments the winning entry has often been one that cannot beat any opponent in direct comparison. It matches cooperative play and punishes defection, so it draws with the decent and loses slightly to the aggressive. It wins the tournament because the games it plays are more productive than the games the aggressive strategies play.

What everyone sees

Competition is read as a series of duels, so the best strategy should be the one that wins duels. That model applies when the payoff is relative and only one opponent exists. In a tournament the payoff is a sum across many encounters, and a strategy that never loses much while enabling large mutual gains outperforms one optimised for defeating others.

What is actually happening

Axelrod ran computer tournaments on the repeated prisoner’s dilemma and reported that the winning strategy never scored more than an opponent in any individual pairing. It was never the first to defect, retaliated when defected against, and forgave quickly, which allowed it to accumulate the payoffs available from sustained cooperation. Maynard Smith’s framework explains why this generalises: what matters is performance against the population actually present, not against a worst case, so the strategic environment is composed of the other players. Beating opponents and scoring highly are different objectives, and a tournament rewards the second.

Why it stays hidden

The result is counterintuitive because we watch pairings rather than totals. In any single match the aggressive strategy visibly does better, and the observer concludes it is stronger. The advantage of the cooperative strategy accrues in the matches where nothing dramatic happens, and undramatic matches are the ones nobody examines.