What the simulation actually does
WARSCAPE is a deterministic, seeded model. One tick is one day. Nothing is decided by a language model: the text explains the numbers, it never produces them. Here is the order the engine works in, every simulated day.
Every day, supplies move first
Each side's depots push ammunition, fuel and spare parts along a network of hubs, rail and roads toward the front. Distance, road quality, damaged hubs and enemy air interdiction all cut how much actually arrives. A sector that receives less than it needs cannot attack for long, no matter how many soldiers stand in it.
Reserves arrive on a schedule, not instantly
Calling up reservists does not put them on the front. They enter a training pipeline with a real delay of three to four weeks and a capacity limit, and arrive at lower quality than regulars. Full mobilization also dips into the wider population, which costs political support at home.
The air war is decided before the ground war
Sorties are generated from operational aircraft, airfield condition and pilot availability, and contested by the other side's fighters and air defence. Whoever holds the air interdicts convoys and supports attacks. Air superiority is not binary: it is a share per sector that shifts day by day.
The front moves where local strength beats local defence
For each sector the engine compares the attacker's effective strength against the defender's, adjusted for terrain, fortification, river crossings, weather, supply and air support, then resolves losses with a Lanchester-style exchange and moves the line. Daily casualties are capped at a realistic share of the force involved.
Wars end politically, not just militarily
Casualties, lost ground, economic damage and time all erode public and political support. When cohesion collapses on one side, or neither side can move for long enough, the model reaches a settlement, a ceasefire or a collapse instead of fighting to the last soldier.
What you control
Who attacks and who defends, whether the United States takes part, how much of the reserve each side calls up, the season, and whether nuclear threats are on the table. Advanced settings add sanctions, strike intensity, weather severity, fog of war, randomness and the victory conditions. Everything else the engine decides from the data.
Questions people ask
Is this a prediction of a real war?
No. It is a model run under assumptions you choose. The same inputs and seed always produce the same war, which makes runs reproducible and comparable, not prophetic. Real conflicts depend on decisions, morale and luck no model captures.
Where do the numbers come from?
Public, country-level sources: the World Bank, SIPRI, the IISS Military Balance and Natural Earth geography, pinned to a versioned snapshot so a run can be reproduced. Every value shows its source and a confidence level on the data sources page. Values not yet tied to a specific dataset record are labelled seed estimates.
How accurate is it?
The engine is calibrated against historical analogues such as the 2008 Georgia war and the opening of the 2022 Ukraine war, and it enforces invariants that rule out impossible states. It is accurate in the sense that its mechanisms behave plausibly, not in the sense that it forecasts outcomes.
What does the model deliberately not do?
No unit positions, no targeting, no route planning, no vulnerabilities, no real-time intelligence, no classified or leaked material. Everything stays strategic and aggregate. Nuclear weapons are either off or an abstract escalation pressure that pushes both sides toward negotiation: there is no nuclear targeting of any kind.
Why does the attacker usually struggle?
Because supply, not manpower, decides most fronts in the model, and attacking costs more of it. Across all 88 land borders in Europe the attacker achieved its objectives in 19 runs, the defender held in 41, and 23 ended frozen.
What is a turning point?
A day the engine flags where momentum shifted durably: a supply collapse, reinforcements landing, air control changing hands. You can branch an alternate timeline from any of them and change one variable to see what would have happened instead.
Watch a prepared war play out day by day, or build your own in about a minute.
Model-generated simulation on public, country-level data. Not a prediction. No operational, classified or real-time information is used.