Forecast skill.Measured in the open.

A shared protocol for model comparisons, prospective records, and reproducible evidence.

Observation coverage3 airport stationsNew York · London · Tokyo

Public baselines1 AI · 2 numerical modelsAIFS / IFS / GFS

Public competitionWeekly model leagueExplore standings and receipts ↗

Model performance

The same valid samples. A transparent comparison.

Archived forecastsLoading date range24 h lead · °C0 / 21 matched cases · 0 days

Matching forecasts and observations

The page will update when the data is ready.

Swipe the table to compare all measures.

A seven-day replay tests the evaluation method. It does not count toward a prospective track record or establish long-term skill.

Daily MAE · All matched stations · °C

AIFSIFSGFS

Matched samples will appear here.