How we collect and verify NWS forecast data.

Official NOAA/NWS point forecasts for daily high temperature. We capture every revision at issuance so you see exactly what the settlement source said, and when.

ASOS/AWOS observations from the same stations NWS uses. Temperatures are paired with forecasts so we can measure error, bias, and hit rate by station and lead time.

Final daily highs from NWS climate products. These are the same values used for settlement on temperature contracts—so our verification aligns with how markets resolve.

Forecasts and observations are ingested on a schedule. Each record is tagged with source (NWS), station, and time. We don't interpolate or substitute other weather APIs.

We compare forecast high to actual high. Actuals use running observations during the day and final CF6 values once the day is closed. Every metric is labeled verified or preliminary.

We compute MAE, bias, hit rate (exact and within 1°F / 2°F), exceedance rates, and lead-time breakdowns. All derived from the same NWS-based pipeline so numbers are consistent.

Temperature prediction markets settle on NWS data. Our inputs are NWS forecasts and NWS-derived observations and climate reports. You're not trading on a different definition of "high" than the one that settles your contract.

Stations and lead times have different accuracy. Our by-station and by-lead-time stats show where the forecast is stable versus noisy. Use that to focus on markets where you have an edge and avoid ones where error variance is high.

MAE, bias, and hit rates give you a clear picture of how wrong the forecast tends to be and in which direction. You can fold that into your probability estimates and position sizing instead of assuming the forecast is unbiased or equally reliable everywhere.