HOW WORLD WAR 3 RISK IS MEASURED

WWIII.REPORT runs its own conflict-scoring system, built here and published in full — formulas, weights, rubrics and calibration set. Every tracked war theater is reassessed every day from open-source reporting.

The scales are calibrated against real wars, not invented cases: 16 historical conflicts from 1914 to 2003 for the theater scale, and 10 whole-world moments for the systemic layer — both world wars, the Cuban Missile Crisis, Korea, Vietnam, the Yom Kippur alert, the Gulf War, Kargil. How the system scores them is shown under tested against history.

A conflict can be catastrophic and move the WW3 risk index barely at all.

That is why there are two, and neither is the total of the other.

A third reading, the conflict-news trend, is counted mechanically from open data on the radar page and each theater. It never touches the two numbers above.

Every theater is assessed for how ready it is to escalate and how strongly it is connected to the others. The global score follows the highest-scoring escalation pathway, not an average — so a severe but contained war can carry a high War Index and still add little here.

Its range names the risk level:

It is built in three steps.

1  How escalated each conflict is

A separate reading from the War Index below, on its own scale. What each input contributes:

2  What links each pair of conflicts

Across 9 theaters that is 36 pairs, each scored on what actually ties the two together. What each link contributes:

3  The system

Each theater then gets two readings: how ready it is to go off — mobilization and failing diplomacy, the things that run hot before the shooting — and how much of the rest of the world travels with it if it does.

If this one ignites, how much comes with it?

The global number is the worst of those, not the average, because a world war needs one place to start rather than a generally tense planet.

One War Index per theater, 0–100: the whole conflict process on one scale, from stable conditions through total war. 0–49 is pre-war escalation; 50–100 is verified active armed conflict; 50 belongs to war alone. Which side of 50 a theater sits on is decided by one fact — whether verified state belligerents are fighting there — never by the number itself. A war in a lull is still a war.

Ground combat, air and missile strikes, territorial change and nuclear posture are graded separately, so a severe air or missile campaign can score highly with no major ground fighting.

Not a probability: a War Index of 42 is not a 42% chance of war. It is a stage on an ordinal scale.

Above 50 — how far the war has gone

The fighting itself sets an internal floor on the war branch, from the last line that matches what is actually happening:

Three dials move that floor by up to 5 points; the result is then mapped onto 50–100, so the old scale’s 45 lands exactly on 50 and 100 stays 100:

Below 50 — movement toward war

A different question, so different inputs — nothing here measures fighting, because there is none to measure. The composite is mapped onto 0–49, so maximum pre-war escalation reads 49 and the outbreak of war can only move a theater up:

We scored ten historical moments on the WW3 risk index — not single wars, but the whole world as it stood at the time. Both world wars rank above every conflict that stayed regional:

Rank the same ten by how much fighting was happening and July 1914 comes last: almost nothing was burning in the month before the largest war in history began. The bloodiest of the ten, Gulf War, 1991, spread nowhere.

Ten scenarios with inputs estimated by us: a test, not a proof, which is why this layer is published as provisional.

AI reads the evidence. Code sets the score.

The model never picks a number. It reads reporting and answers factual questions; a fixed formula does the arithmetic. It is never shown yesterday's score, so today's cannot drift upward by inheriting it.

1. Collect. Open-source reporting from news wires and agencies, once a day.
2. Extract. Each report becomes individual events, graded by how well evidenced they are — confirmed, multiply reported, single source, or disputed.
3. Classify. Each theater is placed on 18 fixed scales, citing sources for every choice.
4. Score. The formulas above turn those into the published numbers.

Participant counts come from the Correlates of War Inter-State War Data v4.0 dataset — 12 wars, 52 distinct countries.

A quiet news cycle is not a de-escalation. Three states look identical in a feed and mean different things:

Conditions that persist — mobilization, deployment, blockade, severed hotlines, nuclear alert — stay in force until there is evidence they ended. A war ends on a verified cessation, never on an empty feed.

Both indexes measure escalation and connection. Neither measures:

• the legality of any action, or who is responsible for it
• which side is winning
• humanitarian severity, casualties or civilian suffering
• a percentage probability, or a prediction that World War III will begin

A conflict causing immense suffering can score lower than a tense standoff with none, because the question asked is how far escalation has travelled and how connected the conflicts are — not how much harm has been done.

GDELT. A machine count of how much of the world's news reports conflict, from the GDELT Project's public event log — a fixed baseline built from 775 sampled days between 2016 and 2026, where 100 is a place's own normal. It has one job here: it can veto a fall. Our four news feeds can go quiet without the fighting stopping, and this count comes from a different corpus, so it can tell the difference. It never raises a score. Data from the GDELT Project.