Skip to content

Repeated Game Recognition and Grim Trigger Analysis

Why It Matters

One-shot game analysis often predicts non-cooperation (Prisoner's Dilemma defection, public goods free-riding), yet we observe extensive cooperation in reality. Repetition fundamentally changes incentives: the threat of future punishment can deter present defection. Grim trigger analysis shows how cooperation can be an equilibrium: the short-term gain from defection is outweighed by the long-term loss from permanent punishment. This explains how cooperation emerges without central enforcement: reputational incentives in ongoing relationships create self-enforcing agreements.

The Key Move

  1. Check if the interaction repeats: same players, observable actions, future matters.
  2. Calculate the one-shot payoffs: what do players gain from mutual cooperation (C,C), mutual defection (D,D), and exploitation (D,C)?
  3. Apply grim trigger logic: "I cooperate until you defect; after any defection, I defect forever."
  4. Calculate defection temptation: immediate gain from defecting while opponent cooperates.
  5. Calculate punishment cost: loss from eternal mutual defection versus cooperation.
  6. Check if future matters enough: if (present value of future cooperation losses) > (immediate defection gain), cooperation is sustainable.
  7. This requires discount factor (how much you value future payoffs) to be sufficiently high.

Classic Application

Maintaining cooperation in repeated Prisoner's Dilemma. One-shot: both defect (dominant strategy). Repeated: grim trigger makes cooperation an equilibrium if players value future enough. Cooperate yields payoff R per round forever (total = R/(1-δ) where δ = discount factor). Defect once yields temptation payoff T, then punishment D forever (total = T + δD/(1-δ)). Cooperation is sustainable when R/(1-δ) > T + δD/(1-δ), which simplifies to δ > (T-R)/(T-D). This inequality shows when repetition enables cooperation: must value future sufficiently and punishment must be costly enough.

Surprising Application

Open-source software contribution. One-shot analysis: free-riding dominates (why contribute when you can use others' code?). Repeated game reality: developers interact repeatedly on platforms like GitHub. Grim trigger equivalent: community members cooperate (contribute code, review, answer questions) with those who've cooperated, but stop helping free-riders (ignore their issues, don't review their PRs). This sustains contribution equilibrium: the immediate cost of contributing is outweighed by long-term benefit of community support. Developers with long-term presence contribute more because they value future cooperation (high discount factor). Non-obvious because there's no formal contract, but reputational repeated game structure creates self-enforcement.

Failure Modes

Assuming infinite repetition - cooperation requires sufficiently long horizon, but most real interactions have uncertain endpoints. Ignoring noise and mistakes - in real settings, you might defect accidentally, triggering permanent punishment. Missing renegotiation - "grim forever" isn't credible when players can restart cooperation. Not checking for discount factor - cooperation requires valuing future enough; short-sighted players still defect. Treating all punishment as grim - many effective strategies use limited punishment (tit-for-tat) rather than permanent defection.

Go Deeper

Axelrod, The Evolution of Cooperation, Chapter 2; Fudenberg & Maskin, "The Folk Theorem in Repeated Games with Discounting," Econometrica 54(3), 1986