ALLIANCES HARDEN. BUDGETS RISE.
RHETORIC FOLLOWS.
The ten signals behind the index sit between 1914 and 1939 on the same scale. Russia–Ukraine is the largest European land war since 1945. Israel–Iran has crossed the precedent of direct exchange. PLA shipbuilding outpaces US Navy procurement. New START expires February 2026 with no successor.
Six figures behind the composite.
One 0-100 score.
Heuristic seed snapshot (will be replaced by daily-refresh GitHub Action). Composite index sits in the upper end of the 'Crisis' band: persistent active conflicts, hardened blocs, record arms spending, decaying arms-control architecture, and several active flashpoints. Lower than late-1939 readings primarily because no single major-power bilateral has crossed the ultimatum/general-mobilization threshold.
What the score is measuring.
Several traditions reading the same data.
Why are great-power conflicts back on the agenda after thirty years of being thought obsolete?
Anarchy among states is the permanent condition; balance of power is the only stabiliser. The post-1991 unipolar moment was always going to end, and great-power competition with China and Russia is simply the system reverting to its long-run baseline. Predicting war by counting power transitions is the discipline's oldest move.
“It is foolish to assume that great-power war is a problem of the past simply because the experience of one generation has not produced it.”
Democracies trade with each other and very rarely fight each other. The institutions built after 1945 (UN, IMF, WTO, EU) made interstate war prohibitively costly for their members. The return of war is therefore a return of failed institutions, not a return of nature, and the fix is institutional repair.
War is not natural; it is industrial, requiring weapons manufacturers, mobilised political consent, and a military-industrial complex with budget incentives to find threats. The 'return of war' is partly the return of arms-industry profits to centre stage and the policy capture they produce.
From Cairo, Delhi, or Pretoria, the binary of 'rules-based order versus authoritarian revisionists' looks like an attempt to draft the South into a Western quarrel. War is not 'back', it has been continuous in Yemen, Sudan, the DRC, Syria; what is new is that it now reaches countries the Western press notices.
“We do not want a new Cold War, and we will not be a battlefield in one.”
From Augustine through Aquinas, the just-war tradition tolerates war only under strict conditions (just cause, last resort, proportionality, immunity for non-combatants). The pacifist strand (Tolstoy, the historic peace churches, large parts of Buddhist thought) rejects even those. Both judge the current wars failing on multiple counts.
From people who have been in wars: the political narratives are reliably the part that survives least. What does survive is the people who came back wrong, the maintenance cost on the people who didn't, and the gap between the speech that started the war and the bodies that ended it. Treat the speech accordingly.
The same signals applied to 1914 and 1939.
Run-up to World War I
Run-up to World War II
Probability by pair, horizon, and severity.
Taiwan-reunification rhetoric is on an explicit timetable; PLA shipbuilding outpaces USN; semiconductor decoupling removes a major peace-incentive; both sides retain off-ramps and economic interpenetration.
Active proxy war with NATO arming one side; Russian doctrine lowered the nuclear threshold; sub-threshold attacks on NATO soil already routine. Direct conventional contact is the most likely accidental escalation path on a short horizon.
Direct missile/drone exchanges already crossed the precedent threshold in 2024–25. Iran's nuclear program degraded but not eliminated; Israel's strike doctrine permits repeat raids; US backstop ambiguous.
Recurring Kashmir-triggered border crises; both sides have demonstrated cross-border strike doctrines; nuclear thresholds publicly low on the Pakistani side.
DPRK ICBM and tactical-nuclear programs mature; sanctions enforcement degraded by Russia partnership; ROK political swings between engagement and hard deterrence.
De-escalation / status quo / escalation.
US – China
Resumed mil-to-mil channels, a Taiwan-Strait crisis-management agreement, and a partial rollback of the most acute export controls. Decoupling continues in critical tech but stops short of total bifurcation.
- · Successful Taiwan election cycle without PLA escalation
- · Restoration of US-China defense hotline use during an incident
- · Reciprocal export-control narrowing
- · Asia-Pacific arms races continue but at slower tempo
- · Forecast 5y 'major' probability ~3%
- · Index drops 5–8 points
Continued gray-zone pressure on Taiwan and the South China Sea, more semiconductor controls, occasional Coast Guard ramming and ADIZ incursions. No bilateral break but no off-ramp either.
- · Routine PLA exercises around Taiwan continue at present cadence
- · Annual tariff/export-control round-trip
- · Two-track diplomacy on climate and fentanyl only
- · Index stable in upper-Crisis band
- · Allied military buildup (Japan, Australia, Philippines) accelerates
- · Increased probability of accidental incident becoming the trigger
PLA blockade or quarantine of Taiwan; US carrier movements challenged. Sub-week decision window; activation of mutual-defense treaties.
- · PLA exercise that converts to actual quarantine
- · Lethal incident at sea with US/Japanese/Filipino casualties
- · Taiwan declaration that crosses Beijing's stated red line
- · Index +20–30 within days
- · Global semiconductor supply collapses
- · Activation of US-Japan / US-Philippines treaty consultations
- · Run on Asian financial centres
NATO – Russia
Front lines frozen along current contact, no formal recognition; partial sanctions relief tied to verifiable troop withdrawals from specified zones; arms-control track restarts (sub-strategic).
- · Mutual exhaustion + leadership succession on one side
- · Third-party (Türkiye / Gulf) brokered framework
- · Visible Russian economy deterioration past tipping point
- · Index drops 8–12 points within months
- · Refugee return begins
- · NATO eastern flank posture remains elevated for years
Attritional war continues; sub-threshold attacks on NATO (sabotage, GPS spoofing, undersea cable cuts, suspicious fires) become routine and tolerated; arms-control architecture fully dies.
- · Continued Western military aid at current pace
- · No succession event on either side
- · New START expires without replacement (Feb 2026)
- · Index stable in upper-Crisis band
- · European defence spending stays at 3%+ of GDP
- · Nuclear-doctrine creep on both sides
Russian munition lands on NATO territory with casualties, or NATO interception of Russian aircraft over Allied airspace with losses. Article 4 invoked; Article 5 debated within hours.
- · Strike on Polish, Romanian, or Baltic infrastructure with casualties
- · Belarusian incursion into Suwalki Gap
- · Tactical nuclear demonstration in Ukraine
- · Index +25–35 within hours
- · Full European mobilization
- · Strategic-systems alert posture rises on both sides
- · Tactical-nuclear use becomes thinkable on a weeks-horizon
Israel – Iran
Iran's program kept below weapons-grade by repeated standoff strikes and sanctions; no formal agreement, but no major direct exchange. Proxy war continues at lower intensity.
- · New IAEA access framework
- · Internal Iranian succession that re-prioritises economy
- · US re-engagement on a JCPOA-successor framework
- · Index drops modestly
- · Saudi-Israeli normalisation completes
- · Proxy fronts (Yemen, Iraq, Syria) continue but at lower tempo
Periodic direct missile/drone exchanges (2024–25 precedent), proxy attacks, and Israeli strikes on Iranian assets in Syria/Iraq/Lebanon. No declared war but an active sub-war state.
- · Continued Iranian enrichment near weapons-grade
- · Houthi attacks on shipping resume
- · Hezbollah rocket campaigns
- · Index stable in Crisis band on this pair's contribution
- · Persistent shipping and energy risk premia
- · Risk of miscalculation crossing into 'major' tier with each cycle
Israeli strike on Iranian weapons-grade facility or leadership; Iranian large-scale missile retaliation including on Gulf states. US drawn in by force-protection requirement.
- · Confirmation of Iranian weapons-grade enrichment
- · Direct attack on Israeli civilian infrastructure with mass casualties
- · Iranian leadership decapitation strike
- · Oil to triple digits within hours
- · US carrier strike group engagement
- · Risk of Iranian breakout under fire (nuclear)
- · Global recession risk
India – Pakistan
Ceasefire on the Line of Control holds; back-channel diplomacy on water and Kashmir continues; no major terror trigger.
- · No major attack inside India attributed to Pakistan-based groups
- · Sustained ISI-RAW backchannel
- · Pair contribution to index drops
- · Bilateral trade resumes modestly
Periodic LoC exchanges; political escalations over water sharing and Kashmir status; no full mobilization.
- · Annual rhetoric cycle around independence days
- · Single-digit-casualty cross-border attacks
- · Index stable; structural risk persists
Mass-casualty attack inside India attributed to Pakistan-based group; Indian cross-border strike doctrine activated; Pakistani retaliation. 2019 Balakot precedent at larger scale.
- · >100-casualty attack on Indian civilians or military
- · Indian leadership domestic-pressure constraint to act
- · Pair contribution +30 in days
- · Public consideration of tactical nuclear use
- · Strategic forces alerted on both sides
Korean Peninsula
Working-level talks resume; family-reunion-tier humanitarian channels; no provocations beyond test cycle.
- · ROK government engagement shift
- · DPRK domestic priority pivot
- · China leverages aid for restraint
- · Modest index drop on this pair
Routine missile tests, occasional artillery shots near Northern Limit Line; cyber and crypto-theft operations continue.
- · Sanctions enforcement remains degraded via Russia partnership
- · DPRK arms exports to Russia continue
- · Index stable on this pair; longer-term arsenal expansion is the bigger story
Naval clash near NLL or DMZ artillery exchange with casualties; ROK proportional-response doctrine triggers escalation chain.
- · DPRK strike on ROK military installation
- · Failed test landing near ROK or Japan
- · Defector-related provocation
- · Index +20 in hours
- · US-ROK Combined Forces Command activation
- · Seoul artillery exposure means cost of escalation is immediate and catastrophic
Sources, weights, and code are open.
Where every number comes from
The composite index is computed from the signals listed on this page, each backed by one or more named sources. Where the source publishes a public dataset or feed it is linked below; where a signal involves qualitative judgement, the LLM-assisted pass is explicitly marked on the signal card.
- ·ACLED
- ·CISA advisories
- ·Crisis Group Tracker
- ·FAS Nuclear Notebook
- ·IISS Military Balance
- ·Mandiant / Microsoft / Google TAG reports
- ·OFAC / EU sanctions registries
- ·SIPRI Arms Transfers
- ·SIPRI Military Expenditure
- ·UCDP
- ·WTO disputes
- ·alliance treaty texts
- ·joint communiqués
- ·national defense ministries
- ·official doctrine documents
- ·official speeches
- ·open-source intel (OSINT) feeds
- ·press conferences
- ·state department releases
- ·state media
- ·tariff filings
- ·treaty depositary records
Everything is versioned
- → Every hourly snapshot is committed to git with a message naming the signals that moved.
- → A daily snapshot is archived to
data/history-current/for the calibration log. - → Raw scraped article lists are written to
data/raw/so a score is reproducible from its input bundle. - → Signal definitions, weights, and seeded scores all live in plain JSON or TypeScript; anyone can open a PR challenging a value and explain why.
How this pillar is scored.
Methodology & limits
Ten signals, each scored 0-10, weighted into a 0-100 composite. The score summarises pressure on the international system. It is not a probability of war. The conflict-pair forecasts are subjective and reissued daily; what they buy is a track record over time.
Pre-nuclear and pre-cyber events are scored on the eight signals that existed then. The composite is normalised over the included signals only, so a 1914 reading and a 2026 reading sit on the same scale within the limits of the data.
Historical events are hand-curated from secondary sources. Current readings are seeded by hand and overwritten hourly by a scheduled scrape plus an LLM-assisted classifier. Every snapshot is in git, so every revision is auditable.