The Threat That Leaves Something to Chance

So far we’ve built threats that work by being believable. A threat you’d actually carry out deters; a threat you’d never carry out is just noise. But that leaves a nasty gap: some responses are so catastrophic that no one believes you’d ever choose them — and yet those are exactly the deterrents you’d most like to point at small provocations. This lesson is about the escape hatch Thomas Schelling found. You don’t threaten the catastrophe. You threaten the risk of it.

The credibility ceiling on big threats

Recall the credibility test from earlier lessons: a threat is credible only if, at the moment of truth, carrying it out is something you’d actually prefer to do (or have committed yourself into having to do). Now try to apply that test to a mismatched threat:

“Nudge one inch across my border and I will incinerate your capital.”

Run the backward-induction check. Suppose the other side does nudge one inch. Do you now launch a nuclear strike — inviting your own annihilation in return — over one inch of dirt? Obviously not. And because they can run that same check, they know you won’t. The threat is empty: too big to be believed. Its very enormity is what kills it.

This is the credibility ceiling: past a certain size, a threat of certain catastrophe becomes less credible the bigger it gets, because the punishment is wildly disproportionate to the provocation and self-destructive to deliver. A speeding ticket is a credible deterrent against speeding precisely because it’s survivable to issue. “Death penalty for jaywalking” deters no one — everyone knows the judge would flinch.

So we have a problem. We possess an enormous deterrent (say, a nuclear arsenal) and a small provocation (a border probe, a blockade run). How do you bring a huge deterrent to bear on a small provocation without the threat collapsing under its own implausibility? Threatening certain disaster won’t work. We need something else.

Brinkmanship, defined

Schelling’s answer is to stop threatening certainty and start manipulating probability. Brinkmanship is the deliberate creation and manipulation of a shared risk of catastrophe that neither side fully controls — Schelling’s own phrase for it is “the threat that leaves something to chance.”

The key move is a substitution. You cannot credibly promise to jump off the cliff — everyone knows you don’t want to die, so the promise is empty (the credibility ceiling again). But you can credibly walk closer to the edge, on a slope that’s slippery, where at any moment you might slip and fall. Walking toward the edge is something you’d actually do, because each step is survivable and reversible-ish — right up until it isn’t.

Brinkmanship, then, is not a single dramatic threat. It’s a process: you raise the shared probability of disaster, hand the other side the choice to keep climbing or back down, and dare them to keep climbing. Every step you take toward the edge is a message — “I am still here, and I will risk more than you will.”

Why leaving it to chance is more credible

Here’s the counterintuitive engine of the whole idea: uncertainty and loss of control become sources of bargaining power.

Compare the two threats one more time, from the opponent’s point of view:

• “I will definitely push the button.” To believe this, they must believe you’d choose mutual annihilation on purpose. You wouldn’t. Unbelievable.
• “Things could get out of hand, and I’m willing to run that risk.” To believe this, they only need to believe you’d tolerate, say, a 5% chance of catastrophe to get your way. That’s entirely plausible — people accept small risks of large losses all the time. Believable.

The probabilistic threat is credible precisely because it doesn’t require you to want the catastrophe. It only requires you to prefer “a somewhat riskier standoff” over “backing down right now.” That’s a bar a rational actor can clear. The certainty threat asks you to be suicidal; the risk threat only asks you to be stubborn.

And there’s a second gift: the risk threat comes with a dial. There’s always “one more turn of the screw.” Raise the alert level. Move the ships closer. Cancel the leave for the reservists. Each notch nudges the shared probability of disaster up a little — visibly, credibly — without ever requiring you to choose the disaster outright. The last person willing to turn the dial wins.

The game of chicken

The cleanest model of all this is a game called chicken. Two drivers race their cars straight at each other. Whoever swerves first is the “chicken” and loses face; whoever holds the line wins. But if neither swerves, both cars crash and both drivers die.

Lay out the payoffs and you can see the whole structure:

Notice the shape. You want to hold — but only if the other swerves. The disaster in the bottom-right corner is so bad that mutual holding is worse than losing. There’s no single “rational” move: what you should do depends entirely on what you think the other will do. That indeterminacy is what makes chicken a coordination-of-nerve game rather than a simple optimization.

Now the terrifying winning move, which ties directly back to the first lesson in this course — commitment. Suppose you conspicuously rip your steering wheel off and throw it out the window where your rival can see it. You have now made it impossible for yourself to swerve. Your rival, watching this, faces a simple updated choice: swerve or die. A rational rival swerves. By destroying your own freedom of action — by tying your hands — you’ve forced the outcome. The player who can credibly commit to not swerving wins.

Brinkmanship is chicken played by inches. Instead of one all-or-nothing wheel-toss, you edge your car forward foot by foot, raising the risk of a crash gradually, forcing your rival to decide at each inch whether this is the moment to peel off. Same logic, finer resolution — and, crucially, you can dial the risk up or down instead of committing to certain collision.

Brinkmanship in the wild

The textbook case is the Cuban Missile Crisis of 1962. The Soviet Union placed nuclear missiles in Cuba; the United States demanded their removal. Here’s the essential point: neither Kennedy nor Khrushchev wanted a nuclear war. Both would have vastly preferred almost any outcome to global catastrophe. So neither could credibly threaten to start one — that would have failed the credibility ceiling instantly.

What they did instead was pure brinkmanship: each raised the shared risk of a war neither wanted, and dared the other to keep climbing.

Brinkmanship isn’t confined to superpowers. The same engine runs at every scale:

• Labour strikes. Neither the union nor the firm wants a long work stoppage — it bleeds both. But each side lets the strike drag on, raising the mutual pain and the risk that the whole relationship breaks down (customers leave, the company folds, jobs vanish for good). The side that can credibly tolerate more of that rising risk wins the better contract. Nobody “threatens” bankruptcy; they just let the pressure build.
• Debt-ceiling standoffs. A government edges toward a default that would hurt everyone, betting the other party will fold before the cliff. Each day without a deal nudges up the chance of an accident nobody wants — a missed payment triggered by a procedural snag rather than a decision.
• Diplomatic and territorial standoffs. Ships shadowing each other, jets intercepting jets, forces massing at a border: no one declares they’ll start a war, but each side raises the ambient chance that a collision or misfire spirals — and dares the other to de-escalate first.

The common thread: in every case, the coercion comes from a manipulated risk of a shared disaster, not from a threat of certain disaster that no one would believe.

Where brinkmanship bites back

Now the catch, and it’s a big one: the risk is real. “Leaving something to chance” means you have genuinely handed part of the outcome to chance — and sometimes the chance fires. The crisis you were only ever bluffing about can become the war you never wanted, through an accident, a miscalculation, a shot-down plane, a misread order. Nobody chooses the catastrophe; the process delivers it anyway.

This is the double edge of the whole strategy. The autonomy of the process — the fact that neither side fully controls it — is exactly what makes the threat credible (you can’t be bluffing about a risk you don’t fully command). But that same autonomy means the risk can go over the edge without anyone deciding to push it there. The feature and the bug are the same mechanism.

The practical lesson is that brinkmanship is a strategy you deploy against your own interest in one sense — every notch you turn also endangers you. It works only when you can tolerate a rising shared risk and keep an exit ramp open, and when the process hasn’t run away from both players at once. Schelling’s admiration for the tactic was always shadowed by his dread of it: the thing that makes the threat believable is the same thing that can get everyone killed.