JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Repeated Games & Cooperation

In a one-shot game, betrayal wins and cooperation dies. But play the same game again and again, and "we'll meet again" becomes a quiet, powerful economic force — enough to make rivals keep their word without any contract at all.

The same game, played twice — and forever

Earlier in this rung the [[prisoners-dilemma|prisoner's dilemma]] delivered a bleak verdict. Two players who would *both* be better off cooperating each find that betraying is their dominant strategy — the best reply no matter what the other does — so they both betray and both end up worse off. That gloomy ending, though, quietly assumed something: that the game is played *once*, by strangers who will never meet again. Now relax that one assumption. What if the same two players sit down to the very same dilemma tomorrow, and the day after, and for as long as either can foresee?

A game played over and over by the same players is a [[repeated-games-tit-for-tat|repeated game]], and it changes everything. In a one-shot encounter, today's betrayal has no tomorrow to answer for it. In a repeated game it does: cheat now and you may pay for it in every round to come. That single difference — the *future hanging over the present* — is what lets cooperation survive a dilemma that destroys it in one shot. Economists call this looming future the shadow of the future, and it is the quiet hero of this whole guide.

Tit-for-tat: be nice, but never a pushover

How exactly does the future discipline a cheater? Through a *strategy* — a rule for what to do each round given what has happened so far. The most celebrated of these is tit-for-tat, and it is almost insultingly simple: in the first round, cooperate; in every round after that, do whatever your opponent did to you last time. Cooperate with cooperators, betray those who betrayed you, and forgive the instant they return to cooperating. That's the whole rule.

In the late 1970s the political scientist Robert Axelrod ran a famous tournament: he invited experts to submit computer strategies that would play the repeated prisoner's dilemma against each other, round after round. Out of dozens of clever, devious entries, the winner was tit-for-tat — the shortest program of all. Axelrod distilled why it did so well into four traits worth memorising: it is *nice* (never the first to betray), *retaliatory* (it punishes betrayal at once, so it is no pushover), *forgiving* (it stops punishing the moment the other cooperates again, so grudges don't spiral), and *clear* (so simple that opponents quickly learn it pays to cooperate with it).

The arithmetic of patience

Cooperation surviving is not magic; it is arithmetic, and it hinges on weighing one fat gain *today* against many smaller losses *later*. This is where game theory borrows a tool from finance you have already met: a future payoff is worth less than the same payoff now, so we shrink — discount — each future round before adding it up. The more you value the future (the more patient you are, and the more likely the game continues), the heavier those future rounds weigh, and the more a one-time cheat costs you.

Let's make it concrete with a tiny dilemma. Each round, mutual cooperation pays each player 3; if you betray a cooperator you grab 5 that round (but then they retaliate forever); mutual betrayal pays only 1 each. Suppose, to keep the future alive, there is a 90% chance the game continues into each next round. Cooperating forever is worth 3 every round, which using the present value of a continuing stream comes to 3 / (1 - 0.9) = 30. Betraying once nets you 5 today, but then you and your victim trade betrayals at 1 forever after — worth 5 + (1 ÷ (1 - 0.9) - 1) × ... roughly 5 then a long tail of 1s, totalling about 14. Thirty beats fourteen, so cooperation wins.

Per-round payoffs:   both cooperate = 3   betray a cooperator = 5
                     both betray   = 1   (the punishment forever)

Keep cooperating forever:   3 + 3(0.9) + 3(0.9^2) + ...  = 3/(1-0.9) = 30
Betray once, punished after: 5 + 1(0.9) + 1(0.9^2) + ...  ~= 5 + 9    = 14

   30  >  14   ->  cooperation is the better deal

Now make the future shaky: chance of continuing = 0.4
Keep cooperating:  3/(1-0.4)            = 5
Betray once:       5 + 1*(0.4/(1-0.4))  ~= 5.7

   5  <  5.7   ->  betrayal now pays.  The shadow of the future shrank.
The same dilemma, two futures. When the game is very likely to continue (90%), the long tail of cooperation (30) dwarfs a quick betrayal (14). Make continuation unlikely (40%) and the maths flips — betrayal pays. Cooperation lives or dies on how heavily the future weighs.

Read off the moral from the two cases. When the future looms large, the lasting reward of cooperating swamps the one-time prize of cheating, so even purely self-interested players keep their word. Shrink the future — make the game likely to end soon, or the players impatient — and the calculation flips: the quick grab wins, and cooperation collapses. Cooperation is not about being kind; it is about the future being valuable enough to protect.

Reputation, retaliation, and the credible threat

What tit-for-tat really wields is a threat: cooperate and I cooperate; cross me and I punish you. But a threat only changes behaviour if your rival believes you will actually carry it out — it must be a [[credible-threat-and-commitment|credible threat]], the idea you met in the guide on sequential games. Repetition is what makes the threat credible. In a one-shot game, vowing "if you betray me I'll punish you" is empty, because there is no later round in which to punish. Repeat the game and the punishment round is real, your willingness to inflict it is on display, and the threat finally bites.

Out of credible threats, repeated across rounds, grows something larger: a reputation. If you have cooperated a hundred times, your next partner trusts you without a contract; if you have a record of punishing cheats, no one risks cheating you. Reputation is, in effect, the future reaching back to govern the present — your past conduct becomes a promise about your future conduct. This is why firms guard their brands, why merchants in a marketplace where everyone returns tend to deal honestly, and why a single betrayal can destroy in a day what took years to build. The threat of lost reputation is often a fiercer enforcer than any court.

This logic illuminates a puzzle from the previous rung. A [[cartel-and-collusion|cartel]] like OPEC is a repeated prisoner's dilemma: every member is tempted to quietly overproduce, yet many cartels hold together for years. How? Because the members meet again and again, watch each other's output, and stand ready to retaliate — flooding the market to punish a cheat — so the shadow of the future keeps each one (mostly) in line. The very same machinery sustains tacit cooperation among a few firms who never sign a thing: each knows that grabbing market share today invites a price war tomorrow.

The last round, and other honest cracks

There is a famous and unsettling catch. Suppose the game is repeated, but everyone knows it ends on a *fixed* final round — say, exactly round 100. In that last round there is no future left to fear, so betraying is the dominant strategy, just as in a one-shot game. But if both players know they will betray in round 100, then round 99 has effectively no future either, so they betray there too — and this backward induction unravels all the way to round 1. The chilling conclusion: cooperation cannot be sustained by pure self-interest in a game with a known last round. It is the open-endedness — not knowing when it ends — that keeps the shadow of the future alive.

Step back and feel how far we have come. The one-shot prisoner's dilemma seemed to prove that rational self-interest dooms us to mutual betrayal. Repetition does not repeal that logic — in the last round it still bites — but it shows that "we'll meet again" can be a real economic force, strong enough to make cooperation self-enforcing without contracts, courts, or kindness. The same idea you started with — a stable outcome where no one wants to change their move, a Nash equilibrium — now has a richer, more hopeful sequel: in a repeated game, *cooperate-and-retaliate* can itself be that stable outcome.