Why does tit-for-tat win Axelrod's tournament?

Tit-for-tat is nice (never defects first), retaliatory (punishes defection immediately), forgiving (returns to cooperation), and clear. It never beats any single opponent head-to-head, but it accumulates the highest total across a round-robin tournament.

Iterated prisoner's dilemma

A round-robin tournament for cooperation.

Six strategies. Every pair plays a repeated game. Selfishness wins each round but loses the tournament — watch Axelrod's nice, retaliatory, forgiving tit-for-tat rise to the top.

Scoreboard · total points

— rounds · — noise

Cooperation timeline

—

Click a strategy in the scoreboard to inspect its matchups.

cooperate defect each bar = one match, left→right = rounds

Strategies in the tournament

Rounds per match

Repetitions of the game200

Noise · mistake rate

Chance a move flips per turn0%

Real interactions are noisy — even cooperators sometimes slip. This is where pure tit-for-tat breaks down and forgiveness pays off.

How to use

Toggle which strategies enter the tournament (at least two). Set rounds per match and a noise %, then run. The scoreboard ranks every strategy by total points across a full round-robin; click any one to see its per-match cooperation timelines below.

Start with noise at 0 to see tit-for-tat dominate. Then push noise to 5–10% and watch forgiving strategies — Pavlov especially — pull ahead.

The strategies

Always Cooperate (AC): Trusting. Cooperates every round.
Always Defect (AD): Selfish. Defects every round.
Tit-for-Tat (TFT): Cooperate first, then copy opponent's previous move. Won Axelrod's 1980 tournament.
Pavlov (Win-Stay, Lose-Shift): Repeat last move if it earned ≥3, else switch. Forgives by re-syncing after noise.
Grim Trigger: Cooperate until opponent defects once, then defect forever.
Random: 50/50 each round. Baseline.

What Axelrod found

In 1980, political scientist Robert Axelrod ran a tournament where game theorists submitted strategies. The winner was the simplest: tit-for-tat (Anatol Rapoport, 4 lines of code).

Axelrod identified four traits of winning strategies:

Nice — never defect first
Retaliatory — punish defection
Forgiving — return to cooperation after retaliation
Clear — predictable, so others can adapt

Selfish defection wins each individual round but loses every tournament. Cooperation, with backbone, dominates over time.

The noise problem

Pure tit-for-tat is fragile against noise. If two TFT players play and one accidentally defects, they enter an endless tit-for-tat retaliation loop.

Add noise (try 5–10%) and watch Pavlov dominate. After an accidental defection it can re-sync to mutual cooperation. Forgiveness is a feature in noisy worlds.

This is one reason humans have apologies, gossip about reputation, and bureaucratic dispute resolution — error-correction layers on top of basic reciprocity.

Citations

Axelrod, R. (1984). The Evolution of Cooperation. Nowak, M. A. (2006). Five Rules for the Evolution of Cooperation. Science.

Built by Joshua Chua · Fable rebuild · vanilla JS + Canvas