The scorecard

How good is it?

We test Snow Day Radar the hard way: hide a whole winter, forecast it blind, then check every school morning against what districts really did. Here is the scorecard, misses included.

15Kentucky winters tested, one at a time
101,632school mornings checked
91%of the time, a closed morning got higher odds than an open one

When it says 60%, do schools close about 60% of the time?

Every morning with winter weather in 15 Kentucky winters (19,888 mornings), grouped by the odds the site would have shown.

80% is our ceiling (never 100%), so mornings at 80% closed even more often than that.Mornings under 20%: schools closed on 6% of them.
When Snow Day Radar said each step, how often schools closed (weather mornings, 15 Kentucky winters)
SaidSchools closedMornings
Under 20%6%7537
20%25%1795
30%37%1219
40%45%976
50%57%817
60%64%827
70%71%882
80%89%5835

Better than counting inches

Compared with a plain “this many inches closes school” rule, its errors are 48% smaller, because it knows what 2 inches means in your district: how rare that snow is there, how hilly the bus routes are, and whether ice or cold is the real problem.

Two and three days out

Fed only the forecast as it stood each evening, it put closed mornings above open ones 95% of the time for the next school day, 93% two days out and 91% three days out. That is why outlooks use bigger steps.

What it still misses

  • Light freezing drizzle that weather stations barely record.
  • Districts that close the day before a big storm arrives.
  • Sleet storms, which snowfall maps undercount. A fix is in testing.
  • South Carolina passed its own test: 18,111 school mornings over 4 winters (2021-22 to 2024-25), each forecast the evening before and checked against the state's missed-day reports. When it said 80%, 89% of those mornings had no in-person school.
  • Everywhere else the model is in beta: it ranks Arkansas closures well (96%) but runs low there, misses Texas ice-storm aftermath, and runs high in snowy northern states like Wisconsin, where a fix is in testing. Each state goes live only after its own test passes.
How the odds are made

Source: the Snow Day Radar engine's backtests (11/7/2026). Closure records: Kentucky Department of Education amended school calendars, 2009-10 to 2025-26; South Carolina Department of Education missed-day reports, 2021-22 to 2024-25.