Humans vs Machines vs Money

Who calls tomorrow's high best?

Explore the national Human vs. the Machine section → High/low temperature ribbons, Ryan’s heat map, and paired NDFD value at each station.

National Weather Service forecasters, the National Blend of Models, the raw global models and the prediction market, all forecasting the same thing: the daily high temperature at U.S. stations, checked against the official settlement value.

The Daily-High Board

mean absolute error in °F by lead time · lower is better · outlined = best at that lead

Each source is scored on every station-day it forecast, so sample sizes differ (hover a cell for n). NWS gridded forecasts stop at Day 7.

Error growth with lead time

°F mean absolute error · hover for every source at that lead

The Money Line

the prediction market as a forecaster: its price-implied expected high, one day out

Next-day high, mean absolute error

Closest on the day

share of station-days each source finished nearest the observed high (ties split)

Week by week

next-day mean absolute error · the market in colour, everyone else in grey · hover for names