How the snow day calculator works, and how accurate it is

Short answer: it reads the National Weather Service forecast for your exact location, scores the snow overnight and before buses run, ice, wind chill and weather alerts against how used to snow your area is, then adjusts for your type of school and caps the result at 95%. How accurate is it? We have not measured that yet: it is a weather-based estimate, not a verified prediction of your district’s decision. Want to try it first? Use the snow day calculator.

By Muhammad Shakeel, who built the method and keeps this page in step with it. Last updated 1 November 2026.

A desk by a snowy window with a weather map, a thermometer, a snow gauge, a wind vane, a magnifying glass and a toy school bus.

How accurate is the snow day calculator?

We have not measured its accuracy against actual school closures yet. A 70% result is the model's estimate of closure risk, not a claim that the calculator has been correct 70% of the time. It also is not the chance of snowfall: weather is one input, while the result estimates disruption to the school day.

The estimate can be wrong even when the weather forecast is right. Our model makes assumptions about how schools respond, and it does not know your district’s policy or actual road conditions. Weather forecasts can also change, so check again before bed and in the morning.

Some sites claim 90% or even 100% accuracy. We have not measured ours yet, so we do not quote a figure. After a full winter of comparing our numbers with real closures, we will publish the results on this page, misses included.

What we do not know

How the estimate is made

Four steps sit behind every result. Nothing is hidden: the result box also lists every forecast number it used.

1. Your location

A ZIP code is turned into a map point using the US Census Bureau's public ZIP code (ZCTA) list. A city name uses the Census list of places. Nothing is sent anywhere for this step.

2. The forecast

We ask the National Weather Service (NWS) for the detailed forecast at that point (its 2.5 km forecast grid). From it we take, for the school day you pick:

We request data when you search. Forecasts are cached for up to 30 minutes and alerts for 5 minutes; your browser may reuse a result for up to 10 minutes. NWS updates may arrive less often.

If forecast hours, weather data or NWS alerts are unavailable, we show the available forecast numbers but leave the percentages unavailable. Missing data does not mean no snow or ice.

For today, the NWS forecast only starts a few hours before now, so snow or ice that fell earlier is not counted, and the page says from what time the forecast starts. Once the school morning is over, we stop estimating today.

3. How used to snow is your area?

The same storm means very different things in different places. We group states into four levels. Places above 1,800 m (about 5,900 ft) move up one level, because mountain towns handle snow better. Low ground in Arizona, California and Nevada, and west of the Cascades in Oregon and Washington (below 1,000 m, about 3,300 ft), moves down one level, because snow there is rare.

LevelSnow that gives about a 50% closure chanceOr iceOr wind chill
Snow is rare here1 in0.03 in10°F
Some snow most winters2.5 in0.06 in-5°F
Snow every winter5 in0.1 in-20°F
Snow belt: schools are well prepared7.5 in0.15 in-30°F

These are model assumptions, not measured district closure thresholds. This is a simple, state-level rule. A snowy corner of a mild state (or the opposite) can be off. For example places in each level and real district rules, read how much snow it takes to cancel school and how cold it has to be.

4. The score

Each factor gives its own chance on a smooth curve: a little snow adds a little, and the chance climbs fast near the model’s baseline snow threshold. Snow during the morning commute counts extra; snow during the school day counts half. Strong wind adds to snow, because it drifts back onto cleared roads. Very low wind chill closes schools on its own in cold states. NWS warnings add weight because forecasters only issue them for serious weather.

The chances are then combined, so two medium risks (say, some snow and some ice) add up to more than either alone. Last, we adjust for the school type:

The delay chance uses the same factors with lower limits (a dusting at 6 am often means a 2-hour delay, not a closure). We show only the part of it that is not already a closure.

We cap every result at 95%. Forecasts change, and districts also weigh things we cannot see, such as staff, buses and how many snow days they have left.

Is there a snow day calculator formula?

There is a scoring formula behind this tool, but no universal formula that tells a school when to close. Our algorithm combines the snow, ice and cold risks described above, adjusts for wind and alerts, then applies the school-type factor. The result panel shows the forecast inputs used for your search. The regional thresholds are assumptions, not a record of how your district voted or a machine-learning model trained on its closures.

Why can two snow day predictors give different results?

They can use different forecasts, time windows and assumptions about schools. Compare when each forecast was retrieved, which location it represents, and whether the percentage means snowfall, a delay or a full closure. A larger number alone does not make a predictor more accurate. Your district's official announcement takes priority.

Changes to the method

Every change to how the estimate is made is listed here with its date, so an older result can be compared fairly.

Ready to check your area? Enter your ZIP code in the snow day calculator.