Axelrod Tournament
Robert Axelrod, 1980: invite game theorists to submit Prisoner's Dilemma strategies. Run a round-robin tournament. The simplest strategy submitted — "Tit for Tat" by Anatol Rapoport — wins. Repeat with a larger, more diverse field: Tit for Tat wins again. Add evolutionary dynamics (strategies reproduce in proportion to score): Tit for Tat and a few cooperative variants dominate. The takeaway: nice, retaliatory, forgiving, clear beats clever defection in any environment containing other reciprocators.
Tit for Tat — the four virtues
Tit for Tat: cooperate on the first round; thereafter, do whatever your opponent did last round.
It satisfies four properties that Axelrod identified as critical:
1. Nice: never defects first.
2. Retaliating: punishes defection immediately.
3. Forgiving: returns to cooperation as soon as opponent does.
4. Clear: easy for opponents to model and adapt to.
A nicer-than-TFT strategy (always cooperate) loses to defectors. A nastier-than-TFT strategy (always defect) loses to cooperators. A less-forgiving strategy (Grim Trigger) does poorly because it locks into mutual defection after a single mistake. A noisier strategy (random) loses to almost everyone.
What happens when noise is introduced: the perfect-information TFT becomes brittle — one accidental defection triggers an endless retaliation loop. The fix: Generous TFT or Tit for Two Tats, which forgive single defections. With noise > 5%, generous variants dominate over strict TFT. This matches biology and human behavior remarkably well — real reciprocity is generous, not strict.
Beyond TFT: in 2012, Press and Dyson identified "zero-determinant" (ZD) strategies that can unilaterally control opponent's score. ZD strategies are exploitative in one-shot but lose in evolutionary settings — they get outcompeted by generous reciprocators. The deep lesson of Axelrod survives.