Live PLATFORM

Live global conflict escalation index, recalculated every two hours.

ww3chance.com translates the density and direction of escalatory geopolitical signals into a single trackable percentage. AI event classification on a fixed scoring rubric, exponential time decay, and a fully published mathematical methodology. Built for traders, analysts, and curious readers who want a defensible signal rather than a vibe.

Live URL ww3chance.com
Sources 4 wire feeds
Classifier Gemini 2.5 Flash
Update cadence Every 2 hours
Event half-life 23 hours
WHAT WE BUILT

A defensible probabilistic signal aggregator.

ww3chance ingests RSS feeds from BBC World, Reuters, Al Jazeera, and Sky News every two hours. Each relevant article is passed to a Gemini 2.5 Flash classifier with a fixed scoring prompt that assigns a conflict zone, event type, direction (escalatory, de-escalatory, neutral), a numerical weight from -2.0 to +3.0, and approximate geographic coordinates. The classifier output is committed to a Cloudflare D1 SQLite database with deduplication by URL hash.

The global index is computed from all events in the last 72 hours using exponential time decay (decay equals weight times e to the negative 0.03 times hours_ago). Physical events score 6.7 times higher than rhetorical events. Per-zone scores are capped to prevent any single conflict from dominating the global signal. The output range is 1.5 percent to 29.5 percent under normal conditions, with values above 30 percent reserved for confirmed nuclear events or direct superpower combat.

TECH STACK
Cloudflare WorkersCloudflare D1Gemini 2.5 FlashLeaflet.jsCloudflare Pages
OPERATIONAL METRICS

By the numbers.

4
Tier-1 wire sources
2h
Recalculation cadence
23h
Event half-life
6.7x
Physical vs rhetorical weight
12
Weight scale anchors
6
Risk band tiers
72h
Active scoring window
100%
Methodology transparency
CAPABILITIES

What the index actually does.

AI event classification with fixed prompt

Every relevant article passes through Gemini 2.5 Flash with a calibrated scoring prompt. Output is structured: zone, event type, direction, weight, coordinates. Reproducible scoring, not vibes.

12-anchor weight calibration

The weight scale is anchored to real-world severity. +3.0 for nuclear use, +1.0 for confirmed cross-border strike, +0.3 for major rhetoric, 0.0 for routine exercises, down to -2.0 for major peace deals.

Exponential time decay

Events decay with a 23-hour half-life. A +1.0 event contributes +0.5 the next day and approximately +0.25 two days later. Stagnant conflicts gradually stop moving the needle.

Per-zone score capping

Physical scores per zone are capped at +/-8, rhetorical at +/-2. No single regional event can dominate the global index.

Six risk band interpretation tiers

0-3% Background Noise, 3-6% Elevated Tension, 6-10% Active Escalation, 10-16% Global Warning, 16-22% War Preparation, 22%+ Critical Threshold. Each tier has a published interpretation.

Comparison to academic benchmarks

The methodology page includes a direct comparison to the Caldara-Iacoviello GPR Index used by central banks. ww3chance measures event severity; GPR measures news coverage volume. Different questions, different tools.

Fully published methodology

Every scoring decision, every formula, every calibration anchor is published at /methodology.html. The system is auditable by anyone who wants to verify a number.

Stated limitations

The methodology page documents what the index cannot do. No classified intelligence access, no black swan prediction, no scientific consensus on weight calibration. Honest limits, published openly.

Live signal feed and zone breakdown

Frontend displays the current index, latest scored signals, 7-day index trend, and per-zone status with 72-hour weighted signals.

WHY THIS PROJECT MATTERS

Why ww3chance is more defensible than its competitors.

Most geopolitical risk indices are either monthly news-coverage counters that lag real events by weeks, or black-box proprietary scores with no published methodology. ww3chance is neither. It updates every two hours, the scoring rubric is published in full, the formula is mathematically reproducible, and the comparison to the academic GPR Index is explicit about what the platform measures and what it does not.

The build sits on Cloudflare D1, which makes the entire scoring backend SQLite-portable. Anyone could clone the platform, point it at different feeds, and re-derive a custom index. That portability is deliberate. The defensibility of the signal comes from methodology transparency, not from infrastructure obscurity.

RELATED WORK

More from the portfolio.

READY WHEN YOU ARE

Want a live signal index like this?

Send a brief. Tileterra Systems responds within 24 hours with scope and pricing scaled to your specific needs.

Start a project