Four feeds, each paired with the instrument it historically relates to and the illustrative historical correlation measured on our own stored data. The effects are modest and sensitive to release-date alignment. The value is the point-in-time data feed, not a trade signal — align it to your own assumptions before relying on any number on this page. This is the free tier; the full catalog and our highest-signal feeds are in the premium API.
Correlation, not causation · not investment advice · free-tier illustrative.
These are historical correlations measured on stored data, not forward paper-trading claims — see /signal-live for the forward test of one of these feeds.
Several of these correlations weaken materially when aligned to the date the data was actually publishedrather than the date it refers to — e.g. EU fuel prices → UNG: t≈4 at the reference date falls to t≈1.1 at the actual Thursday publication date, a lookahead artifact. The original figures remain below with corrections noted. Publication-date alignment is the reason the underlying product is a point-in-time feed.
EU weekly retail fuel prices (150k+ rows, official EC source)
Corrected: +1.10 full-sample / +0.15 OOS, aligned to actual Thursday publish date
European energy-complex pricing leads the US nat-gas ETF at a weekly horizon.
Corrected 2026-07-16: the t above was measured against the bulletin's Monday reference date, entered Tuesday — but the bulletin is not published until Thursday. Aligned to the actual publication date, t falls to ≈1.1 (OOS ≈0.15). A publication-lookahead artifact.
LME nickel warehouse tonnes
Base-metals inventory dynamics co-move with the copper-mining complex.
t below 2 — a genuine tilt, not proof-grade.
BLS Job Openings & Labor Turnover
Labor-demand momentum leads broad equities.
High IC but weak t, reflecting a short effective sample — high IC does not imply significance. BLS publishes JOLTS ~5 weeks after the reference month, so the same release-date caveat as EU-oil→UNG applies; treat as illustrative only.
NAAIM active-manager equity exposure survey
Stretched manager positioning tends to precede weaker forward SPY — a classic sentiment fade.
Weakest of the four (|t|<1); included as a representative of the sentiment/positioning category.
These four were selected because in-sample and out-of-sample IC agree in sign — no sign-flips, no fitted lag selection. The effects are modest. The premium product is the data: the full point-in-time feed catalog, plus the highest-signal feeds we hold back (compute / AI-adoption). No signal has cleared our full validation process to date.
Complete re-verification results — including candidates that did not pass — are in the Research Lab. Request access to the full feed catalog at /api.