Live2026-09-26 05:40 UTC
Home › Data sources

Where every number comes from

engine v4.5.1 · all sources public

Every input is a public source anyone can reach: 41 news feeds, open market-price services, a public central-bank statistics service, a public crypto data API and the encyclopaedia’s pageview counts. No paid wire, no private data, nothing behind a login. This page lists, index by index, what feeds it and how much.

No third-party index sits inside any of our numbers. We buy nobody’s risk score and republish nobody’s sentiment gauge. Where an outside index exists for a similar idea, we build our own from raw inputs so no provider can switch our numbers off or change them under us.

Index by index

shares are the published weights

World Tension Index

  • Our own news corpus — 60%. conflict, trade and media factors are all corpus measurements
  • Oil and gas prices — 20%. the energy factor
  • Government bond yields — 8%. part of the finance factor
  • Equity volatility — 8%. part of the finance factor
  • Precious metal prices — 4%. part of the finance factor
  • Public attention — overlay. used in the shock and media components

Energy Security Index

  • Oil and gas prices — 85%. oil 40, gas 30, volatility 15
  • Precious metal prices — 15%. the gold-to-oil ratio
  • Our own news corpus — context. supply-threat coverage feeds the shock overlay

Food Security Index

  • Grain prices — 100%. wheat, corn and rice, each ranked against its own history
  • Our own news corpus — context. export bans, drought and fertiliser coverage

Currency Stress Index

  • Currency data — 100%. dollar index plus the major pairs, each scored on its own
  • Government bond yields — context. rate decisions shape the reading

Shipping Stress Index

  • Dry-bulk freight rates — 70%. the core reading — flagged "est." when the print is unavailable
  • Our own news corpus — 30%. chokepoint, port and route coverage from the corpus

Sanctions Pressure Index

  • Our own news corpus — 80%. share of world coverage about new measures and enforcement
  • Currency data — 20%. stress in sanctioned currencies

Gold & Silver Fear Index

  • Precious metal prices — 70%. both metals against their own recent ranges
  • Government bond yields — 15%. real yields
  • Currency data — 10%. the dollar
  • Equity volatility — 5%. market volatility

Crypto Fear & Greed Index

  • Crypto market data — 85%. price momentum, 30-day volatility, dominance
  • Our own news corpus — 15%. fearful crypto coverage in the corpus

AI Risk Index

  • Our own news corpus — 80%. corpus share 40, severity of the strongest headlines 40
  • Public attention — 20%. attention on AI-safety topics against three years

Inflation Pressure Index

  • Other WTM indexes — 100%. energy 35, food 25, shipping 20, currency 20

Supply Chain Risk Index

  • Other WTM indexes — 100%. freight 40, energy 25, sanctions 20, conflict 15

Market Fear Index

  • Other WTM indexes — 100%. gold & silver fear 50, crypto sentiment inverted 50

The inputs, in plain language

InputHow it is collected
Our own news corpusAll 41 public news feeds, read in full twice a day. Every headline is scored for tension, tagged with the countries and conflicts it names, and — since engine v4.5 — grouped with its duplicates so one story counts once.
Precious metal pricesDaily settlement prices for gold and silver from a public market-data service, with a second independent service as fallback.
Oil and gas pricesBrent crude and natural gas daily closes, same two-service arrangement.
Grain pricesWheat, corn and rice daily closes.
Currency dataThe dollar index and the major pairs, daily closes.
Government bond yieldsThe US 10-year yield from a public central-bank statistics service, with a market source as fallback.
Equity volatilityThe standard US equity volatility index, daily close.
Dry-bulk freight ratesA published dry-bulk shipping rate index. This is the one input that is not always available; when it is missing the last value is carried forward and every page that uses it is flagged "est.".
Crypto market dataBitcoin price, 30-day volatility and market dominance from a public crypto data API.
Public attentionDaily pageview counts for 21 topic pages on the public encyclopaedia, ranked against three years of their own history.
Social pulseTwo public world-news discussion feeds, used only as a minor cross-check.
Other WTM indexesNo new data: a fixed weighted average of indexes this engine already computes.

We name source types rather than commercial vendors: vendors change, and a number that depends on one named supplier is a number somebody else controls. Every one of our market inputs has a second independent service behind it, and the news feeds are listed by name on the evidence page.

How often, and what happens when a source fails

Every 10 minutes: market prices, crypto data, all derived indexes, every page and the API.

Twice a day (00:10 and 12:10 UTC): the full news corpus, the conflict roster, country scores, AI Risk and its sub-indexes, public attention.

Every night: monthly stories, timelines, datasets, sitemaps.

When a source is down: the last known value is carried forward and marked estimated, every page using it shows “est.”, and the field is listed in /api/v1/status. Scores are never quietly filled with a guess. If a news feed dies, the corpus is smaller that cycle and the confidence reading falls, which is exactly what it is for.

When the method itself changes: it ships as a dated release in the changelog, the engine version on every page moves, and historical values are never recomputed.