DEMFI Demo

Before choosing a city, check which ones have the highest forecast confidence. This matrix shows our model's reliability for each city and forecast horizon (D+1 to D+5). Focus on cities marked HIGH (dark green) — that's where DEMFI's statistical edge is strongest. Moscow is highlighted as the demo city.

DEMFI aggregates hourly temperature forecasts from 14 numerical weather prediction (NWP) models (ECMWF, GFS, ICON, JMA, and more). Each line is one model's forecast trajectory; ensemble models also show their spread (p10–p90). The black line is actual observations. V0 shows the raw model forecasts; V1 applies DEMFI's Markov bias correction. Both are included in the monthly all-cities plan.

Using all forecast models, DEMFI computes the probability that tomorrow's maximum temperature falls in each bucket. The tallest bar is the most likely outcome. V0 is built from the raw model forecasts; V1 applies DEMFI's Markov bias correction. Both are included in the monthly all-cities plan.

The Portfolio shows Edge NO positions only: the buckets that Polymarket overprices according to DEMFI. Buying NO means betting that this bucket will not happen. Stakes are sized with the Kelly Criterion (quarter-Kelly, capped at 25%). Included in the monthly all-cities plan.