It's April 2027, China Is Massing Off Taiwan, and You Have Minutes to Decide

We turned the defining scenario of a US–China war into a free browser wargame calibrated to CSIS's 2023 study — and it grades you like the Pentagon. Here's what playing it teaches.

It's 6:00 a.m. Greenwich time — 06:00 Zulu, on the military's universal clock — April 15, 2027. A FLASH cable, the highest-urgency message the system can send, hits your desk from the Joint Staff's J2, the Pentagon's intelligence directorate:

China's People's Liberation Army launches a record-scale "joint exercise" east of Fujian: 42 warships, 4 amphibious assault ships, 150+ air sorties, and — alarmingly — requisitioned civilian RO-RO ferries, the roll-on/roll-off kind you can drive a tank onto. CIA reads it as posturing. DIA, the Pentagon's own intelligence agency, can't rule out invasion prep. The intelligence is ambiguous, the stakes existential.

You are the United States. Not a general, not a pundit — the decision-maker in the room. And the first thing the simulation does is take away the one thing you want most: certainty.

That's Taiwan Strait Crisis, a strategic simulation we built for Go Bananas. It runs in a single browser tab (free to play — no download, no signup: play it now). Under the hood it's a fairly serious piece of work: its parameters are calibrated to "The First Battle of the Next War," the 2023 wargame study from CSIS — the Center for Strategic and International Studies, a Washington defense think tank.

The setup: you have the levers, not the answers

You manage the crisis across four tracks — military, diplomatic, economic, and intelligence — choosing from 45 distinct actions: surge surveillance (the satellites, drones, and spy planes the military calls ISR), open a presidential hotline to Beijing, reposition carrier strike groups, give a prime-time address. And six of those 45 sit on a fifth track the interface won't even show you at the start: nuclear. They unlock only when the crisis reaches DEFCON 2 — and by then you'll wish they hadn't.

Behind the buttons sit systems named Escalation, Nuclear Risk, Strait Control, and Domestic Support — plus nine hidden ones, from CCP internal dissent to PLA logistics strain to semiconductor panic, that stay invisible until your intelligence picture grows sharp enough to hint at them — or the crisis forces them into view.

A cascade engine of 45 rules makes sure nothing you do is isolated. Take one button: Carrier Forward Deployment. Move the strike group toward the Strait and you've bought real deterrence — and put 6,000 sailors inside the engagement envelope of China's DF-21D "carrier-killer" ballistic missile. Escalation ticks up; oil traders price in a fear premium. Hardliners in Beijing get louder; at home, support slips a few points, and hidden congressional pressure starts to build. One click; half a dozen systems move, some you can't yet see.

Eight AI actors — China, Taiwan, Japan, Australia, India, the Philippines, South Korea, and the EU — pursue their own interests around you. They are not props: South Korea may declare neutrality and bar US bases from Taiwan operations, the Philippines can pull its navy from the South China Sea ("We cannot fight China alone"), and India might mobilize 200,000 troops on China's border — helping you enormously, for reasons that have nothing to do with you.

And then there's the clock.

The mechanic that makes it honest: a DEFCON clock that eats your time

This is the design decision we're proudest of, because it captures something most armchair strategy misses.

The simulation tracks DEFCON — the US military's five-step alert ladder, where 5 is peacetime and 1 means nuclear war is imminent — using the real codewords: FADE OUT (5), DOUBLE TAKE (4), ROUND HOUSE (3), FAST PACE (2), COCKED PISTOL (1). As the crisis escalates, two things shrink at once:

• Time per turn collapses from 24 hours at DEFCON 5 to 30 minutes at DEFCON 1.
• Actions per turn drop from four down to one.

Read that again. The deeper the crisis gets, the less you can do and the less time you have to do it. Escalation doesn't just raise the stakes — it amputates your options. The strategists call this "maneuver space," and watching it close around you is the single most instructive feeling the game produces. Every move you wish you'd made at DEFCON 4 is a move you can no longer afford at DEFCON 2.

▶ Feel the clock close in yourself — free, in your browser →

The fog is the point

Most strategy games give you a god's-eye map. This one gives you an intelligence feed that lies to you the way real intelligence does.

Reports arrive with confidence tiers. Two credible human sources — spies — contradict each other. A National Intelligence Estimate lands with the agencies split — CIA pegs invasion odds at 15–25%, DIA at 55–70% — and the Director of National Intelligence's note to the President is brutal: "You must act on incomplete information — there is no option to wait for certainty."

And the feed doesn't just blur — it can lie. Drop below a confidence threshold and the simulation will sometimes silently swap a pair of contradictory reports — disinformation slipped into your own feed, with no tell that it happened. Meanwhile a miscalculation engine rolls fourteen scripted accidents, their odds loading up as the alert level climbs: Taiwanese radar misreads an airliner as a missile wave and sirens sound across Taipei; a US destroyer fires on an unidentified boat in fog — a Taiwanese fishing crew; a cable network airs unverified footage of a nuclear launch and retracts it twenty minutes too late. Wars start by accident. This one can too.

Worse, the picture decays. The simulation models PLA communications adaptation: as the crisis unfolds, the adversary changes how it talks and your sensors go dim — a direct nod to Millennium Challenge 2002, the infamous wargame where Red Team commander Paul Van Riper went radio-silent and gutted a fleet that couldn't see him coming (we built a playable simulation of that one too). Your information advantage is perishable, and a thinking enemy will spend the war making it worse.

Forecasting the unknowable: the six futures

Before you commit to anything, the game plays the rest of the crisis out 200 times — same position, chance falling differently each run — and counts how often each ending comes up. (Statisticians call it Monte Carlo simulation — the technique behind hurricane-track cones and election forecasts.) The result: a probability fan of six futures. At the opening baseline:

• Negotiated Settlement — 25%
• Conventional Victory — 20%
• Successful Deterrence — 15%
• Pyrrhic Victory — 15%
• Taiwan Falls — 15%
• Nuclear Exchange — 10%

Sit with that distribution. The single most likely outcome of a Taiwan war isn't a clean American victory — it's a negotiated settlement nobody loves. Deterrence succeeding and Taiwan falling outright are equally probable. And before you've touched a single lever, the forecast already puts a one-in-ten chance on nuclear exchange. Play well and you can grind that number toward zero; misjudge one signal and you'll watch it climb, turn by turn, in real time.

Eight ways it ends — and why none of them are "you win"

The crisis resolves into one of eight end states: successful deterrence, conventional victory, negotiated settlement, nuclear ceasefire, Taiwan falls, domestic collapse, stalemate, and nuclear exchange. If deterrence fails, the invasion itself is simulated wave by wave — a CSIS-calibrated crossing clock, roughly 18% of China's force per wave, six waves, about a day apart.

Two of those endings deserve special attention.

Domestic collapse. You can lose this war without losing another battle — by losing the home front. When the news confirms the first American combat deaths since Afghanistan and the President stays silent, the question becomes political: how many American lives is Taiwan worth? If Domestic Support craters, Congress revolts and your mandate evaporates. The simulation treats public will as a strategic resource you can spend to zero.

Nuclear use. The game implements a nuclear "ladder" with a hard ceiling — taboo → rung → cap, in the design notes. Crossing the threshold doesn't unlock a power move; it caps your possible score no matter what follows. The game refuses to let nuclear use ever be "worth it." That's not squeamishness — it's the most defensible thing the simulation asserts: there is no clean way to climb that ladder back down.

The lessons, stated plainly

Here's what playthrough after playthrough hammers home — and what the game itself will tell you in its after-action report:

1. You will decide blind. The hardest call — invasion prep or mere coercion? — must be made before the evidence is conclusive, because waiting for certainty is a decision, and usually the losing one.

1. Escalation is a trap door, not a dial. Up is easy and fast; down is slow, costly, and sometimes impossible. You felt this the first time the DEFCON clock shrank your turn.

1. The ammunition runs out first. Munitions run dry in days, not months — calibrated to CSIS's finding that key long-range anti-ship missiles are exhausted in 7–10 days of high-intensity combat. The winner is often the side whose industrial base can keep shooting. Logistics, not heroics.

1. The home front is a front. Public will is finite and spendable. A militarily "winnable" war can be politically un-survivable, and the enemy knows it.

1. Interdependence is both shield and hostage. TSMC — the Taiwanese company that makes most of the world's most advanced chips, including the one in your phone — deters an invasion and becomes its highest-value target. In one branch, Taiwan detonates its own 3nm fabs (chip factories) rather than let them fall, shattering global chip supply for 5–8 years: economic fallout that dwarfs the war itself.

1. There is no good way to use a nuke. The taboo is load-bearing. Break it and even a "win" is ash.

1. The realistic outcomes are the grey ones. Deterrence that holds, a settlement nobody celebrates, a frozen stalemate. The clean, decisive victory is, statistically, the exception — not the plan.

None of this is a prediction. It's a posture — a way of thinking under uncertainty that you internalize far faster by making the calls yourself than by reading another op-ed.

Then the game grades you

You don't have to take our word for it: every run ends with an after-action report the Pentagon would recognize — a letter grade from A — Strategic Masterstroke down to F — Catastrophic Failure, scored across five weighted categories (Taiwan sovereignty 25%, casualties minimized 25%, nuclear outcome 20%, alliance intact 15%, economic damage 15%), plus lessons learned specific to how your crisis ended. Then comes the global leaderboard: your personal bests alongside every other player who's sat in the chair. Playing needs no account; log in and your score joins the cross-player board.

Why this lives on Go Bananas

Taiwan Strait Crisis — Monte Carlo engine, eight AI actors, global leaderboard and all — is a single HTML file running in a sandboxed browser tab, playable on a phone. We built this one by hand, as a showcase of what a Go Bananas game can be (right down to that leaderboard, a platform feature every game gets for free). The platform's real trick runs the other way: you type a description, and about half a minute later you're playing a game an AI wrote for you — then you refine it in chat ("make the timer more forgiving," "add a hurricane on turn five"), with every version saved. Here's the full manifesto.

And if you think our escalation model is wrong — too hawkish, too gentle on Beijing, too kind to the carriers? Remix it. One click forks any published game into a private draft of your own, free, with automatic, permanent attribution. Change the DEFCON timings. Rewrite the cascade rules. Publish your version and defend it on the leaderboard.

If a wargame can live in a browser tab, so can your idea — try: "A Cuban Missile Crisis simulator where I play Khrushchev, with an intelligence feed that's sometimes wrong, five possible endings, and a clock that speeds up as tensions rise."

You have minutes. The President is asking for options.

*▶ Play Taiwan Strait Crisis now — free, no signup: gobananas.co/game/taiwan-crisis*

▶ Make your own at gobananas.co — describe it in a sentence, play it in about half a minute. A free account comes with 150K AI tokens: roughly 10–20 games on the house.

Describe a game in plain language; get a real, playable, shareable game back; refine it by chatting.