Weather markets resolve on a physical station reading, not a panel of judges. That makes them the cleanest test of the same engine that powers Sports Intelligence: price versus what actually happens, across 51 cities.
Five global forecast models rarely agree on tomorrow's high within a degree. When the market's implied bin drifts from where that consensus — and the station history for that city — actually points, the gap is measurable, and it closes the moment the reading comes in.
No opinions involved — the market settles on a thermometer.
ECMWF, GFS, ICON, UKMO and GEM — five independent global models, plus live station readings, per city.
Where the five models agree — and where they don't — sets a confidence band around the likely reading.
Compare that band to the market's implied bin — when they disagree, the edge is the distance between them.
Two resolved signals, both linked to the market they were called on.
Wins and losses, both — the ledger updates every 6 hours.
Every number here is recomputed from the public ledger, not a one-off marketing claim. View public proof →
Free members see the format. Pro members see the price.