How to Measure Geopolitical Risk: Inside the WW3Risk Index

“Geopolitical risk” is often discussed as if it were self-evident. In practice, measuring it requires choices: which signals count, how they are normalised, how missing data are treated and what happens when an exceptional event breaks the assumptions of an ordinary continuous model.

WW3Risk is built to make those choices visible.

The design goal

The index is not intended to estimate the probability of a future war. It measures observed international strategic conditions associated with escalation toward major-power conflict and wider war.

The public score runs from 0 to 100 and is calculated only when required inputs are valid for the same observation date.

Signal 1: GPR Threat

GPR Threat captures the acute geopolitical-threat environment. In the current provisional architecture it is the largest standalone continuous component and has a coefficient of 0.65.

For the latest complete observation, the GPR input is approximately 90.61 and contributes about 57.33 index points.

Signal 2: Strategic Militarisation

Strategic Militarisation provides a slower structural backdrop using complete-case P5 military-burden evidence. Its role is intentionally smaller and slower-moving than the acute threat signal.

The latest structural input is based on the 2024 observation and contributes approximately 2.65 points to the current complete reading.

Signal 3: Direct Conflict

Direct Conflict captures realised interstate violence using UCDP-based calibration. Covered zero remains zero; missing coverage remains missing.

That last rule is important. If conflict data are not yet available for a newer date, WW3Risk does not treat the missing value as peace and does not carry the last conflict observation forward silently.

The co-elevation interaction

The model adds an explicit interaction between GPR Threat and Direct Conflict. The logic is that simultaneous high geopolitical threat and high realised conflict represent a different strategic environment from either signal in isolation.

The interaction is not a causal attribution. It does not say the conflict caused the threat reading or vice versa.

The current provisional formula

m = Militarisation / 100
g = GPR / 100
d = Direct Conflict / 100

b = 0.10m
a = 0.65g + 0.10d + 0.25gd
Continuous = 100 × [b + (1-b)a]

The latest complete inputs reproduce the public 83.90522 reading.

Why rare events use floors instead of points

Some strategic discontinuities are too important to force into arbitrary cardinal scores. WW3Risk therefore keeps categorical nuclear states and rare scenario predicates separate from the ordinary arithmetic.

Two production floors are currently approved: 80 for confirmed direct nuclear-power combat and 96 for confirmed nuclear use. A floor sets a minimum; it does not add bonus points.

Freshness is part of the model

A risk index can be numerically precise and still be misleading if its source dates are hidden. WW3Risk publishes freshness explicitly. The latest complete continuous observation is 31 August 2026 because Direct Conflict is not yet covered for a newer same-date calculation.

The public state therefore says LATEST_ONLY rather than pretending the latest complete score is a current-day value.

Why this matters for reproducibility

The model, public scale, source roles, scenario semantics and publication hashes are versioned. The website receives approved output through a signed publication bridge, so the public presentation layer does not independently invent or edit the analytical state.

That architecture is designed to make later methodological changes traceable rather than silent.

Explore the full methodology, compare the historical series, or download the October 2026 Strategic Risk Report.