What this index is
The WW3 Chance Index is a weighted probabilistic signal aggregator. It ingests geopolitical news from major international sources every 2 hours, classifies each event using an AI model, assigns a numerical escalation weight, and computes a composite probability score.
It does not predict the future. It measures the current density and direction of escalatory signals across active global conflict zones — and translates that into a single, trackable number.
Data sources
The index ingests RSS feeds from the following sources, checked every 2 hours:
Articles are deduplicated by URL hash. Only articles relevant to military conflict, diplomatic relations, weapons, or geopolitical tension are processed. Irrelevant articles (sports, entertainment, domestic politics) are discarded.
Event classification
Each relevant article is passed to an AI classifier (Gemini 2.5 Flash) with a fixed scoring prompt. The classifier assigns:
- Zone — which conflict region the event belongs to
- Event type — military action, diplomatic, rhetorical, nuclear, proxy, economic
- Direction — escalatory, de-escalatory, or neutral
- Weight — a numerical value from −2.0 to +3.0
- Coordinates — approximate geolocation for the map
Weight calibration
The weight scale is anchored to real-world severity. These are fixed reference points used to train consistent classification:
| Weight | Event type |
|---|---|
| +3.0 | Nuclear weapon used, or direct credible nuclear threat with military posture |
| +2.0 | Direct strike on sovereign territory by a major power (US, Russia, China) |
| +1.5 | Major chokepoint closure (Hormuz, Bosphorus) or carrier group deployment |
| +1.0 | Confirmed airstrike, missile launch, or cross-border military incursion |
| +0.5 | Major military mobilisation or formal war declaration |
| +0.3 | Significant escalatory rhetoric from head of state or defence minister |
| +0.1 | Minor proxy skirmish, localised incident, unconfirmed reports |
| 0.0 | Routine exercise, diplomatic statement, or unverified claim |
| −0.3 | Ceasefire announcement or peace talks initiated |
| −0.5 | Confirmed diplomatic backchannel or mediator engagement |
| −1.0 | Troop withdrawal, military stand-down, or formal agreement signed |
| −2.0 | Major peace deal or conflict resolution |
Index calculation
The global index is computed from all events in the last 72 hours. Recent events are weighted more heavily using exponential time decay:
physicalScore[zone] = Σ decay(physical events) — capped at ±8
rhetoricalScore[zone] = Σ decay(rhetorical events × 0.15) — capped at ±2
zoneScore[zone] = physicalScore + rhetoricalScore
rawScore = Σ all zoneScores
compressed = √max(rawScore, 0) × 4.2
probability = 1.5 + min(compressed, hardCap)
The half-life of any event is approximately 23 hours — meaning an event scored at +1.0 will contribute +0.5 to the index the following day, and approximately +0.25 two days later.
The output is mapped to a range of 1.5% to 29.5% under normal conditions. Values above 30% are reserved for confirmed nuclear events or direct superpower military engagement — thresholds that have never been triggered in the post-Cold War era.
What the numbers mean
| Range | Interpretation |
|---|---|
| 0–3% | Background noise. Normal geopolitical tension, no active escalation vectors. |
| 3–6% | Elevated tension. Escalatory signals in one zone, diplomatic pressure rising. |
| 6–10% | Active escalation. Military posturing across multiple zones, superpower rhetoric escalating. |
| 10–16% | Global warning. Concurrent crises with direct superpower involvement. High alert. |
| 16–22% | War preparation. Major military mobilisation, alliance activations, chokepoints threatened. |
| 22%+ | Critical threshold. Reserved for nuclear events, direct superpower combat, or multi-front war. |
How this differs from the Caldara-Iacoviello GPR Index
The Caldara-Iacoviello Geopolitical Risk Index (published in the American Economic Review, 2022) is the most widely cited academic geopolitical risk measure, used by central banks, the IMF, and academic researchers. It is built by counting how frequently war-related keywords appear across 10 major newspapers, expressed as a share of total articles that month. It updates monthly.
The WW3 Chance Index takes a deliberately different approach:
| Feature | WW3 Chance Index | Caldara-Iacoviello GPR |
|---|---|---|
| Update frequency | Every 2 hours | Monthly |
| What it measures | Strategic event severity | News coverage volume |
| Method | AI event classification + weighted scoring | Keyword frequency in newspapers |
| Distinguishes rhetoric vs action | Yes — physical events score 6.7× higher | No — a speech and a strike count equally |
| Source credibility weighting | Yes — NGO claims capped, wire services full weight | No |
| Output | Probability-style percentage | Dimensionless index value |
| Historical dataset | Launched March 2026 | Back to 1900 |
The key conceptual difference: the GPR measures how much geopolitical tension is in the news. The WW3 Chance Index measures how severe the underlying events actually are. A prolonged regional war generates sustained high GPR readings simply because it stays in the news — regardless of whether the conflict is escalating or stagnating. The WW3 index decays events over 72 hours and weights physical military actions far above rhetoric, so a conflict that has been ongoing for months at the same intensity gradually stops moving the needle.
Limitations
This index has deliberate limitations you should understand:
- It cannot access classified intelligence or diplomatic backchannels
- It relies on public news sources, which may lag real events by hours
- AI classification is probabilistic — individual events may be miscategorised
- Black swan events by definition cannot be anticipated by any model
- The weight calibration reflects editorial judgment, not scientific consensus