Change the Game, Not the Players

Five lessons in, you can find the resting point of almost any strategic tangle: spot each player’s best response, locate the cell where nobody wants to move alone, and call it the equilibrium. Powerful. But there’s a trap hiding in that power, and it’s an emotional one. The moment you find an equilibrium you hate — everyone defecting, a race to the bottom, a commons stripped bare — the instinct is to blame the people. “If they’d just cooperate!” “If they weren’t so greedy!” That instinct is a dead end, and this final teaching lesson is about the move that actually works instead.

Here’s the reframe. A Nash equilibrium is simply what a system of self-interested players settles into given the payoffs they face. It’s not a moral verdict on the players; it’s the mechanical resting point of the incentives. So if you dislike where the ball rolls to, you don’t lecture the ball — you retilt the table. You change the payoffs so that the behaviour you want becomes each player’s best response. Do that, and the new behaviour holds itself in place, no sermon required.

This is where two other mental models plug straight in. The incentives model says people respond to the rewards and penalties they actually face, not the ones you wish they’d care about. The leverage points model says some interventions barely nudge a system while others move it wholesale — and the highest-leverage intervention of all is rewiring the rules and payoffs that generate the behaviour, rather than fighting the behaviour itself. Changing the game is that intervention, in its purest form.

Watch an equilibrium move

Let’s make it concrete with the game you already know cold — the prisoner’s dilemma. Mutual cooperation pays (3, 3), but each player is tempted to defect for 5 (leaving the sucker with 0), and mutual defection pays a grim (1, 1). The unique equilibrium is (Defect, Defect): both players, acting perfectly rationally, march straight into the outcome that’s worse for both of them. No amount of “please cooperate” fixes it, because defecting is a best response no matter what the other does.

Now stop asking the players to be better and start editing the table. Use the steppers to subtract a penalty from both players’ Defect payoffs — imagine an enforceable fine of 3 that lands on anyone who defects. Drop the lone-defector’s 5s down to 2, and the mutual-defection 1s down to −2. Watch the NE badge jump.

Nothing about the players changed. They’re exactly as self-interested as before. What changed is the cost of defection, and once defecting stopped paying, cooperation became the rational choice all on its own — self-enforcing, needing no further nagging. That is the entire art of changing the game: find the behaviour you want, and make it a best response.

The levers that change a game

“Change the payoffs” sounds abstract until you notice it’s how half of civilisation’s rules were designed. Here are the four big families of levers, each of which is really just a way of editing the matrix that people face.

Contracts and enforcement

A binding agreement with penalties turns a promise into a payoff. Two players who’d both love to cooperate but each fear being the sucker can sign a contract: defect and you pay a penalty big enough that defection stops being worth it — exactly the fine you just used on the matrix. Enforcement is the machinery that makes the penalty real, and without it the contract is just a nicer-sounding sermon.

This lever cuts both ways, which is why the law watches it. Cartels are illegal precisely because a binding contract would work — rival firms would love to sign an agreement to hold prices high, sustaining the collusive equilibrium that’s terrible for consumers. Ban the enforceable contract and the firms slide back into competing (defecting on each other), which is the whole point. The same lever, pointed the other way, is a peace treaty with verification: each side would break the truce for advantage, so you add inspections and penalties that make breaking it cost more than keeping it.

Law, regulation, taxes, and subsidies

When a government wants to move an equilibrium, it edits payoffs by decree. A Pigouvian tax on pollution adds the cost of the harm back onto the polluter’s payoff, so the privately rational choice lines up with the socially good one — the factory that was rationally dumping now rationally cleans up. Fishing quotas attack the tragedy of the commons head-on: left alone, every boat’s best response is to overfish (grab the fish before a rival does) until the stock collapses; a quota rewrites the payoff of that extra haul into a fine, and suddenly restraint is the best response. Subsidies run the same play in reverse, paying people to do the thing you want until doing it becomes their best move. Every one of these is the penalty-on-the-matrix trick wearing a policy hat — and it’s the direct answer to the commons and externality dilemmas.

Repetition and reputation

You don’t always need a contract or a regulator — sometimes you just need a future. Play a dilemma once and defection wins. Play it over and over with the same partner, and the payoffs quietly change: a defection today buys you a short-term gain but costs you all the cooperation that partner would have offered tomorrow. When that shadow of the future is long enough, cooperating becomes the self-interested move, which is how cooperation emerges among purely self-interested players with not a saint in sight. Reputation is the same mechanism spread across a crowd: cheat one person and everyone hears, so the cost of defecting balloons far beyond the single deal. (The repeated-games lesson in Game Theory Basics is the full treatment — but the headline is that repetition is a payoff-changing lever, not a separate magic.)

Changing the players or the information

Sometimes the cheapest edit isn’t to a payoff but to who’s at the table or what they can see. Add a regulator or a referee and you’ve inserted a player whose whole job is to punish defection. Make moves observable — publish the prices, audit the books, put the vote on record — and secret defection becomes impossible, which changes its payoff to “certain punishment.” Escrow lets two distrustful strangers trade by parking the stakes with a neutral third party until both deliver. Hostages or bonds — a deposit you forfeit if you cheat — are the ancient version: put something you value on the line, and your own incentive to defect evaporates.

Where the model lies

You now have a genuinely sharp tool. Before you go swinging it at everything, here’s the safety briefing — because every model is a lie that’s usefully wrong, and the skill is knowing exactly where Nash equilibrium stops being trustworthy. Five places it bites.

1. Equilibria need not be efficient or fair

The single most important caveat, and the one the prisoner’s dilemma tattooed on you: stable does not mean good. (Defect, Defect) is a rock-solid equilibrium and a small disaster for everyone in it. An equilibrium is just where self-interest comes to rest — it carries no promise of being efficient, fair, or desirable. So never let “well, it’s the equilibrium” function as a justification. It’s an explanation of why a bad outcome persists, and often a to-do list for what payoffs to change — but it is not evidence that the outcome is acceptable.

2. Real people don’t always reach the equilibrium

The theory assumes flawless, calculating players. Real ones are boundedly rational: they miscalculate, they learn slowly, they follow habits and hunches, and they make honest mistakes. Sometimes they land on the equilibrium only after playing for a while, and sometimes they never quite get there. Worse, when a game has multiple equilibria, the theory underdetermines the outcome — it tells you the set of resting points but not which one a given group will actually pick. That gap is equilibrium selection, and the maths alone can’t close it; you need focal points, history, communication, or norms to predict where real players land.

3. Nash ≠ dominant strategy

Two ideas that feel similar and aren’t. A dominant strategy is a move that’s best no matter what anyone else does — a luxury that only some games hand you. A Nash equilibrium exists far more broadly (every finite game has at least one, possibly in mixed strategies), but a player’s equilibrium move is only best given what everyone else is doing, not in general. So don’t assume a dominant strategy exists — most games don’t have one — and don’t read “this is a Nash equilibrium” as “this is the obviously best move for me to make regardless.” It’s the best move conditional on the equilibrium holding, which is a much weaker, more fragile claim.

4. One-shot vs repeated

The very same game gives opposite predictions depending on whether it’s played once or many times. Analyse a long-running relationship — a supplier you’ll deal with for years — as a cold one-shot dilemma, and you’ll wrongly predict betrayal, because you threw away the future that actually sustains cooperation. Run the error in reverse — treat a genuine one-off, no-tomorrow encounter as if reputation and retaliation were on the line — and you’ll wrongly expect cooperation that has nothing holding it up. Before you solve a game, ask how many times it’s really played and whether the players remember. Get that wrong and every downstream prediction is wrong.

5. Payoffs are a model

The whole analysis rests on a quiet, load-bearing assumption: that you know everyone’s payoffs. You wrote down the matrix. But real people value things you probably didn’t put in the boxes — fairness, spite, reputation, loyalty, identity, the sheer pleasure of not being pushed around. In lab experiments people routinely reject profitable-but-unfair deals and cooperate where the “rational” matrix says defect, because their real payoffs include feelings your matrix ignored. And here’s the sharp edge: if the true payoffs differ from the ones you assumed, the true equilibrium can differ too. A game is only as trustworthy as the payoffs you fed it — model the wrong rewards, and you’ll predict the wrong resting point with total, misplaced confidence.

Recap

You’ve walked the whole course. Here’s a mixed quiz that reaches back across every lesson — best responses, the definition, the prisoner’s dilemma and Pareto efficiency, coordination and focal points, mixed strategies, and this lesson’s lever. Six questions; no penalties, just a final tune-up before the exam.

When to use it

Reach for this move whenever you want to change a strategic outcome and the players won’t budge on their own — which is most of the time, because they’re not being stubborn, they’re being rational. Don’t argue with the equilibrium; design around it. Ask: what behaviour do I want, and how do I make that behaviour a best response for each player? Then pull a lever — a contract with teeth, a tax or subsidy, a repeated relationship, a referee, an observable move, an escrow, a bond. When you get it right, the outcome you wanted becomes self-enforcing: it holds because everyone’s own interest now holds it there, and you never have to nag again. That’s the difference between wishing a system were different and actually changing it.

Where this goes next

That’s the course. Six lessons ago, “equilibrium” was a word from an economics headline; now you can find the resting point of a price war, a treaty, a traffic pattern, or a boardroom standoff — spot best responses, name the equilibrium, tell efficient from merely stable, coordinate on a focal point, mix when you must, and, as of today, move an equilibrium you don’t like by rewiring what its players are paid. You can even say where the model lies, which is the mark of someone who understands a tool rather than just wielding it.

One thing stands between you and the certificate: the Final Exam. Fair warning — it plays by different rules than the friendly practice quizzes. It runs one question at a time, and once you submit an answer it locks for good: no Back button, no retry, no Restart. Your score appears only at the end, and you need 70% to pass. It’s the rigour you’d want from anyone who claims to understand Nash equilibrium rather than just having read about it. You’ve done the work across six lessons — trust the questions you’ve learned to ask, then go take it.