Back to Dashboard

Index Methodology

How the WW3 Risk Index score is calculated, calibrated, and updated in real time.

What We Measure

The WW3 Risk Index measures the probability of a direct military conflict between at least two major nuclear powers — specifically the United States, Russia, and China — escalating to or constituting a World War 3 scenario. It does not measure the severity of any ongoing regional war.

Regional wars (e.g. Russia–Ukraine, Israel–Iran) are factored in only to the degree they risk drawing major nuclear powers into direct confrontation.

Score Composition

Each refresh cycle the index is computed in two steps: first a structural baseline is established from standing geopolitical conditions, then each incoming headline applies a micro-delta to that baseline.

Formula
S = clamp(B + Σδᵢ, 0, 100)
S = final index score (0.0 – 100.0)
B = structural baseline (~40 pts as of early 2026)
δᵢ = per-article float delta assigned by AI model
clamp = result is bounded to [0, 100]
Structural Baseline (early 2026)
Russia–Ukraine war ongoing, stalemated+6.0
Israel–Iran direct exchanges, no major escalation+4.0
US–China tensions over Taiwan, no incidents+3.0
North Korea nuclear program active+2.0
Multiple active regional conflicts (Sudan, Myanmar, Sahel)+3.0
US military presence in Middle East at elevated levels+2.0
Breakdown of key arms control treaties (New START suspended)+4.0
Active US–Russia ceasefire diplomacy−3.0
No NATO Article 5 trigger events−2.0
Structural baseline≈ 40.0

The baseline is updated with each model analysis based on the model's assessment of the standing geopolitical situation, separate from individual news events.

Per-Article Delta Scale

Each headline receives a float delta. The scale is intentionally conservative — most real-world events produce small changes. Large deltas are reserved for events that materially alter the probability of direct nuclear-power confrontation.

DeltaEvent Class
±0.1Routine — statement, sanction, minor skirmish
±0.2Significant — notable military movement or breakdown
±0.5Major regional — large strike, credible nuclear rhetoric
±1.0Proxy escalation — direct casualties involving major-power assets
±2.0Near-miss — accidental clash between nuclear-power forces
±5.0Direct conflict — first confirmed shots between nuclear powers
±10.0War — full-scale attack on nuclear-power territory or nuclear use

Calibration Scale

The index is calibrated against documented historical crisis moments to prevent score inflation:

ScoreLevelDescription
0–10LOWBaseline peacetime. Normal diplomatic tensions.
11–20LOWElevated rhetoric, sanctions, no active major conflict.
21–35ELEVATEDActive proxy wars. Nuclear threats in rhetoric only.
36–50ELEVATEDProxy wars escalating, direct incidents between major powers.
51–65HIGHDirect military clash between major powers. Casualties confirmed.
66–80HIGHLimited direct warfare between nuclear powers. Tactical nuclear alert.
81–95CRITICALFull-scale war between 2+ nuclear powers. Weapons on alert.
96–100CRITICALGlobal nuclear exchange in progress.

Data Sources

News is aggregated in real time from 24 RSS feeds across major international outlets, wire services, and conflict-focused publishers. Sources are refreshed every 30 seconds. No social media data is used.

BBCCNNSky NewsNYTGuardianWaPoReutersNPRAl JazeeraJ PostTimes IsraelAl ArabiyaFrance24DWEuronewsPolitico EUNHKSCMPDefense NewsCBSABCKyiv IndependentMeduza

AI Analysis

Analysis is performed by GPT-4o-mini acting as a senior geopolitical risk analyst. The model is explicitly instructed to:

  • Establish a structural baseline from standing geopolitical conditions before reading any headlines.
  • Apply micro-float deltas (e.g. +0.1, +0.3) for routine events — not integer jumps.
  • Only assign large deltas (≥2.0) for events that directly involve confirmed military contact between nuclear-power forces.
  • Treat regional proxy wars (Ukraine, Gaza, Iran–Israel) as already priced into the baseline — their headlines adjust the score minimally unless they draw in US/NATO or China directly.
  • Compare current signals against historical benchmarks: Cuba 1962 = ~75, Able Archer 1983 = ~58, Kosovo 1999 = ~38, Ukraine invasion Feb 2022 = ~32.
  • Deduplicate similar headlines — only the first occurrence earns a delta, subsequent similar articles score 0.
  • Round the final score to one decimal place.

Update Frequency

News fetch
Every 30s
RSS feeds polled continuously
Risk analysis
Every 30s
Full AI re-evaluation each cycle
Chart history
Per refresh
Score appended on each analysis
Disclaimer: The WW3 Risk Index is an experimental, AI-generated metric for informational purposes only. It does not constitute intelligence analysis, financial advice, or an official geopolitical assessment. Scores reflect machine-learning interpretation of public news and may be inaccurate or delayed.