SNOWMADS Today's forecasts →

Receipts, not vibes

The Scorecard

Every night, our forecast for the next day is archived before the weather happens — then graded against verified snowfall the morning after. We can't retro-edit it, and we publish the misses next to the wins. No other snow forecaster shows you this.

64nights scored
69resorts tracked
3.4cmavg error, storm days
74.3cmbiggest dump — Treble Cone

Notable calls

The forecasts that mattered — including the ones we got wrong.

The phantom storm

Mt Buller · 2026-07-12

Verified win
Snowmads≈4cm — too warm, falling as sleet
OpenSnow23–33cm
Mt Buller ski patrol"another 3cm" — 24cm natural base, +1.5°C

A big system hit Victoria, but Buller's elevation sat near the freezing line. Our freezing-level handling read it as sleet; the resort's own patrol report settled it the next morning.

The one we under-called

Treble Cone · 2026-07-13

Our miss
Snowmads23.7cm
Verified actual32.6cm

A real NZ dump that our model blend averaged down — the single strongest input had it right at 40cm and we pulled it toward quieter models. Logged, and it feeds how we weight models next. Publishing the misses is the whole point of this page.

Three forecasts, one storm

Treble Cone & Cardrona · 2026-07-14

Verified win
SnowmadsTreble Cone 13cm · Cardrona 1.6cm (next 2 days)
OpenSnowTreble Cone 15.2cm · Cardrona 10.2cm
Snow-ForecastTreble Cone 36cm · Cardrona 17cm
Verified actualTreble Cone ≈5cm · Cardrona ≈0.5cm

All three issued the morning of Jul 14 for the same mid-mountain elevations. The storm largely fizzled — Treble Cone got about 5cm, Cardrona barely half a centimetre. Both competitors over-called it hard (Snow-Forecast worst, at 36cm and 17cm); our restrained numbers were closest on both mountains. The whole point of the freezing-level blend is not chasing storms that don't land.

Night by night

One-day-ahead forecast vs verified actual, most recent first.

NightResorts scoredAvg error*Biggest event
2026-09-1232quiet everywhere (correctly called)
2026-09-1169quiet everywhere (correctly called)
2026-09-10692.8cmRoundhill — forecast 16.2cm, got 22.3cm
2026-09-09691.4cmTreble Cone — forecast 9.1cm, got 12.4cm
2026-09-0869quiet everywhere (correctly called)
2026-09-07693.5cmTūroa — forecast 2.7cm, got 7.1cm
2026-09-066910.0cmRoundhill — forecast 16.6cm, got 37.1cm
2026-09-05691.9cmTreble Cone — forecast 6.7cm, got 10.1cm
2026-09-04693.9cmTreble Cone — forecast 12.4cm, got 23.1cm
2026-09-03693.3cmquiet everywhere (correctly called)
2026-09-02692.3cmŌhau — forecast 0.7cm, got 1.5cm
2026-09-01693.0cmTūroa — forecast 6cm, got 8.8cm
2026-08-31831.1cmThe Remarkables — forecast 8.7cm, got 9.7cm
2026-08-30551.7cmSunshine Village — forecast 0.1cm, got 5.5cm

*Mean absolute error on “signal” rows — nights and resorts where either the forecast or the actual was ≥2cm. Quiet days everyone gets right are excluded so the number stays honest.

The storm ledger

Every verified double-digit day since we started keeping score, and what we said the night before.

DateResortWe saidIt gotError
2026-09-10Mt Dobson13.6cm11.3cm+2.3cm
2026-09-10Roundhill16.2cm22.3cm-6.1cm
2026-09-10Ōhau12.5cm13.9cm-1.4cm
2026-09-10Mt Hutt10cm12.2cm-2.2cm
2026-09-10Treble Cone14.7cm11cm+3.7cm
2026-09-10Cardrona14.1cm13.1cm+1cm
2026-09-09Treble Cone9.1cm12.4cm-3.3cm
2026-09-06Mt Dobson11.8cm25.1cm-13.3cm
2026-09-06Roundhill16.6cm37.1cm-20.5cm
2026-09-06Ōhau11.7cm16.4cm-4.7cm
2026-09-06Mt Hutt11.9cm23.7cm-11.8cm
2026-09-06Treble Cone6.1cm10.4cm-4.3cm
2026-09-05Treble Cone6.7cm10.1cm-3.4cm
2026-09-04Roundhill6.7cm12.8cm-6.1cm
2026-09-04Ōhau6.4cm11.1cm-4.7cm
2026-09-04Treble Cone12.4cm23.1cm-10.7cm
2026-08-27Mt Dobson21.3cm23.7cm-2.4cm
2026-08-27Roundhill24.6cm39.7cm-15.1cm
2026-08-27Ōhau27.2cm37cm-9.8cm
2026-08-27Mt Hutt19.6cm27.1cm-7.5cm

How this works

Forecasts blend four global weather models — ECMWF, GFS, ICON and JMA — with a temperature-based snow-to-liquid correction, so cold dry powder isn't under-counted the way fixed-ratio forecasts do. “Verified actual” is the weather model's own next-day analysis of what fell, corrected the same way, and cross-checked against official resort snow reports where resorts publish them. It's an honest yardstick, not a marketing one: mountains with good stake data sometimes record more than the analysis shows. Scoring runs nightly and this page rebuilds automatically — last updated 2026-09-12.

See today's forecasts — 69 resorts