How the Snow Day Chance Formula Works

How does Snowday calculate a snow day chance? It converts 48 hours of hourly Open-Meteo forecast data into a 0–100 score whose core is a logistic curve centered on each state’s closure baseline — for example, 6 inches in Michigan — with snow weighted 1.5× during the 5–9 a.m. commute and bonuses for cold below 15 °F, wind above 20 mph, and existing snow depth. The full formula is documented below.

The data source

The bottom line: two free, keyless public APIs from Open-Meteo, called from your browser — nothing passes through any server of ours.

Geocoding: geocoding-api.open-meteo.com/v1/search converts your ZIP code or city into coordinates. Results are filtered to US locations (country code US).
Forecast: api.open-meteo.com/v1/forecast returns hourly snowfall, snow depth, temperature (2 m), wind speed (10 m), and precipitation probability for the next 48 hours, plus daily summaries, with timezone=auto so all window math happens in your local time. Open-Meteo blends government models including NOAA and ECMWF inputs; its data is published under CC BY 4.0.

Metric units are requested and converted client-side (cm → inches at 2.54, °C → °F, km/h → mph) so there is no unit ambiguity anywhere in the pipeline.

The formula, weight by weight

The bottom line: a logistic base curve for snowfall, three capped weather bonuses, then three profile multipliers — every constant is in the table below and in the source code comments.

“Tomorrow” means the next calendar day at your location. The overnight window runs 6 p.m. today to 5 a.m. tomorrow; the commute window runs 5–9 a.m. tomorrow, when buses roll and roads are still unplowed.

Snow day chance formula weights
Factor	Rule	Weight
Base snowfall curve	Logistic: `100 / (1 + e^(−1.1 × (inches − baseline)))`	Sets everything; 50 points at the baseline
State closure baseline	Inches ≈ a coin-flip closure, by state	VT/ME/NH 8 · AK 9 · MI 6 · NY 6 · CO/UT/MN 5 · WA 3 · default 4
Commute-hour snow	Snow falling 5–9 a.m.	×1.5 vs overnight snow
Forecast confidence	Average precipitation probability (clamped 35–100%)	Scales snow totals linearly
Cold	Hours below 15 °F (blowing/drifting risk)	+0.8/hr, cap +10
Wind	Hours at or above 20 mph	+0.8/hr, cap +10
Existing snow depth	Depth already on the ground	+0.5/in, cap +6
School level	Elementary / middle / high	×1.12 / ×1.04 / ×0.96
School type	Public / private	×1.00 / ×0.92
District strictness	Lenient / typical / strict	×0.82 / ×1.00 / ×1.18

The final score is clamped to 0–97 and rounded to a whole percent. The 97% cap is deliberate honesty, not a bug.

Accuracy — the honest part

The bottom line: the score is good at ranking mornings (big-storm vs quiet) and deliberately humble about the exact number, because two things we cannot see make the final call.

Forecasts themselves are uncertain. A 6-inch forecast can verify as 2 or 10. Our probability weighting helps, but it cannot remove model error.
Closures are human decisions. Bus driver availability, road-crew timing, and ice on one bad hill can close a district that our score rates 30% — or keep one open at 80%.

So treat the number as a smart guess, not a guarantee. If it says 70% and your district stays open, the formula was wrong about a human being — that will keep happening, and we would rather tell you that than pretend otherwise.

Known limitations

The bottom line: no delays, no ice-specific modeling, no district-level tuning — and the score is only as good as tomorrow’s forecast.

Two-hour delays are not predicted; the score covers full closures only.
Freezing rain is modeled only indirectly (via temperature), even though ice often closes schools faster than snow.
District strictness is a slider, not a database — we do not track individual districts.
State baselines are single statewide values; a storm can behave very differently 100 miles apart (lake-effect snowbands are the classic case).
US locations only. The baselines mean nothing in, say, Québec, where norms differ entirely.

Last updated: October 2026