Nothing is hidden. Every retaliation, every mobilization, every silence in the simulation comes from the decision tree below — probabilities estimated from publicly available doctrines, treaties, and strategic literature. Randomness only enters at dice-roll time: the rules themselves are deterministic.
Source: the simulator’s escalation engine (escalation.js / constants.js). These diagrams are a faithful projection of the code that actually runs.
This engine models political and military escalation, not real-world prediction. Human leadership, diplomacy, accidents, intelligence failures, and domestic politics enter only as probabilistic rules.
The engine processes retaliation round by round — internally, a breadth-first search over the alliance graph: each round’s strikes generate the next round’s retaliations, up to 10 rounds or until nobody fires anymore.
The heart of the simulator. When a country is struck, each of its allies walks this tree. The “phase” sets how fast and how willingly they act: a NATO member reacts in minutes, an SCO member on the other side of the world reacts in days — and usually only condemns.
The struck country itself needs no alliance to strike back. A nuclear strike on a nuclear power triggers near-certain retaliation — a baseline of 85% that only a prior cyber attack on command-and-control (C3I) can pull lower. That is mutually assured destruction.
Automatic retaliation system (inspired by the Soviet “Perimetr”): if a nuclear power equipped with a Dead Hand is struck before it has fired, retaliation launches on its own — even if command has been decapitated. In the simulation, only Russia has one (Perimetr — the only automated dead-hand system widely documented in public).
Base probability that an ally retaliates for a struck partner, depending on the treaty binding them. An “ironclad” guarantee (USA → Israel, Russia → Belarus) is unconditional; the SCO is only a dialogue forum, not a military alliance. The figures are the engine’s calibrated design parameters, not measured frequencies.
| Alliance | Nuclear retaliation | Conventional retaliation |
|---|---|---|
| “Ironclad” guarantee (USA→Israel, Russia→Belarus) | 100 % | 90 % |
| NATO / OTAN (art. 5) | 70 % | 55 % |
| CSTO / OTSC | 50 % | 30 % |
| Bilateral treaty | 40 % | 30 % |
| GCC / CCG | 35 % | 20 % |
| Axis of Resistance | 15 % | 80 % |
| SCO / OCS | 10 % | 10 % |
| EU mutual defense (art. 42.7) | 0 % | 30 % |
| AUKUS | 0 % | 0 % |
Modifiers: each ally already committed raises the pressure (+12% per ally, bandwagon effect, capped at 95%); a P5 power (USA, Russia, China, France, UK — the five permanent UN Security Council members) whose direct ally is nuked rises to at least 90%; a prior C3I cyber attack removes 25 points; an ongoing conventional war multiplies by 1.3.
When the cascade ends, the outcome is determined by who fired, at whom, and with what total tonnage (in megatons). The thresholds are simplified abstractions inspired by nuclear-winter research (Robock et al., 2007).
| Outcome | Condition (first line reached) |
|---|---|
| EXTINCTION | ≥ 3 P5 powers fired nuclear, OR ≥ 4 nuclear powers fired, OR ≥ 440 Mt — total nuclear winter. |
| WORLD WAR | ≥ 2 great powers (P5) struck each other, OR ≥ 100 Mt. |
| LIMITED EXCHANGE | At least one nuclear weapon detonated — the long-standing nuclear taboo has been broken, and there is no clean “armistice” anymore. |
| ARMISTICE | No nuclear weapon used. Deterrence held. |
The tonnage thresholds are drawn from the nuclear-winter literature: 440 Mt is the standard full US–Russia exchange modelled by Toon et al. — 4,400 warheads at 100 kt, producing roughly 150 teragrams of soot — and 100 Mt sits below it, where the soot loading turns hemispheric. Across 1,620 simulated wars the median war detonates 17 Mt and the heaviest 839 Mt, so both lines are genuinely crossed: they reclassify 93 wars that the head-count criteria alone would have judged less severe.
NUKEMAP, by historian Alex Wellerstein, shows what one detonation does to the place it hits: blast, heat, fallout and casualties around ground zero. Dystopian World starts where that stops. It models who retaliates, through which alliance, and how a single strike becomes a contained exchange, a world war or a nuclear winter. The two answer different questions, and they work well side by side.
It depends on the treaty that binds the struck country to its allies. The table in section 05 gives the base odds the engine uses: an “ironclad” guarantee (the US to Israel, Russia to Belarus) answers every time, NATO’s article 5 about 70 % of the time with nuclear weapons, a bilateral treaty such as the US one with Japan or South Korea 40 %, the SCO almost never. One rule overrides the table: a permanent Security Council member whose direct ally is hit by a nuclear weapon answers at least 90 % of the time. Allies already at war raise the pressure on the others, which is how one strike pulls in dozens of countries.
No. The probabilities are design parameters calibrated on public doctrines (SIPRI, official statements), and every run draws them again, so the same matchup can end in an armistice or in extinction. The scenario pages show how often each outcome comes up over many runs of the same engine.
Yes. It runs in the browser, in English and French, with no account and no download.