Skip to main content

Snowfall scorecard

How close was the snow forecast?

Over the 22 days to Oct 29, 2026, the forecast SnowSure published missed the day's snowfall by 4.4 cm on an average graded day, and by 10 cm or more on 10% of days. It was the closest forecast on 33% of days.

Forecast Trust — Verified, not guessed — corpus scale, blend vs best model, and the full record.

Average miss

4.4 cm

per graded day

Big misses

10%

days off by 10 cm or more

Most often best

33%

days closest of every forecast

On average it called 1.0 cm too much snow a day. Graded on 3,599 resort-days at 90 resorts.

Early season: graded since Oct 8, 2026, over few storms and with the ground still warm. These figures will move as winter fills in.

Every forecast, days 1–14

ForecastAverage missBig missesMost often bestGraded days
NBM (US blend)3.7 cm6%35%2,794 · 67 resorts
SnowSure forecast4.4 cm10%33%3,599 · 90 resorts
ICON (Germany)5.8 cm18%21%1,033 · 65 resorts
Météo-France5.8 cm16%22%423 · 52 resorts
ECMWF5.9 cm17%30%3,002 · 66 resorts
GEM (Canada)5.9 cm16%23%1,603 · 74 resorts
HRRR (US, 2 days)6.0 cm17%20%167 · 51 resorts
NAM (US, 2 days)6.1 cm17%22%155 · 50 resorts
UK Met Office6.2 cm18%21%819 · 55 resorts
JMA (Japan)6.9 cm22%18%2,050 · 72 resorts
GFS7.1 cm23%24%2,550 · 98 resorts

2 forecasts have fewer than 50 graded days here and are not shown.

Head to head, on the same days

Models run over different parts of the world and to different horizons, so their own averages are not like for like. On the days both were graded, here is how often each was closer.

ModelShared daysSnowSure closerModel closerTied
ECMWF2,60662%25%13%
GFS2,33666%20%14%
NBM (US blend)2,09149%44%7%
JMA (Japan)1,78667%24%9%
GEM (Canada)1,41060%30%10%
ICON (Germany)94759%29%12%
UK Met Office74257%30%13%
Météo-France40552%32%16%
HRRR (US, 2 days)15159%27%14%
NAM (US, 2 days)13559%24%16%

By how far ahead

Days 1–3

ForecastAverage missBig missesMost often bestGraded days
NBM (US blend)4.4 cm9%29%479 · 57 resorts
SnowSure forecast4.5 cm9%28%613 · 59 resorts
ECMWF5.3 cm16%24%603 · 59 resorts
ICON (Germany)5.5 cm15%23%480 · 53 resorts
Météo-France5.8 cm16%22%423 · 52 resorts
HRRR (US, 2 days)6.0 cm17%20%167 · 51 resorts
UK Met Office6.0 cm16%21%462 · 55 resorts
NAM (US, 2 days)6.1 cm17%22%155 · 50 resorts
GEM (Canada)6.3 cm17%24%532 · 57 resorts
GFS6.3 cm17%21%532 · 57 resorts
JMA (Japan)7.0 cm22%18%548 · 58 resorts

2 forecasts have fewer than 50 graded days here and are not shown.

Days 4–7

ForecastAverage missBig missesMost often bestGraded days
NBM (US blend)3.8 cm6%32%1,149 · 64 resorts
SnowSure forecast4.0 cm8%31%1,192 · 78 resorts
GEM (Canada)5.8 cm16%23%819 · 70 resorts
ECMWF6.0 cm17%21%978 · 65 resorts
ICON (Germany)6.1 cm20%19%553 · 63 resorts
UK Met Office6.3 cm21%21%357 · 51 resorts
GFS6.7 cm22%15%749 · 82 resorts
JMA (Japan)7.0 cm22%17%868 · 63 resorts

1 forecast has fewer than 50 graded days here and is not shown.

Days 8–14

ForecastAverage missBig missesMost often bestGraded days
NBM (US blend)3.4 cm5%42%1,166 · 55 resorts
SnowSure forecast4.7 cm11%36%1,794 · 84 resorts
GEM (Canada)5.5 cm14%20%252 · 64 resorts
ECMWF6.0 cm18%41%1,421 · 61 resorts
JMA (Japan)6.8 cm22%18%634 · 67 resorts
GFS7.7 cm25%33%1,269 · 90 resorts

How it is graded

  • The SnowSure forecast is the one we published, day by day, as it appeared on the site and in the app — not a blend rebuilt afterwards.
  • Truth is measured snow: SNOTEL, national weather services' stations and other gauges, from a station standing within 300m of the forecast's height. A model's own analysis never grades a forecast, and a depth gain does not either, since settling makes it less than what fell.
  • A day counts when snow fell or was forecast (at least 1 cm either way), so a false alarm costs. Dry days nobody forecast are left out.
  • Most often best compares every forecast graded for the same resort, day and issue, where there were at least three; a tie counts for each.
  • Forecasts with fewer than 50 graded days are not shown. The window is the last 90 days, recomputed daily. Updated Oct 30, 2026.

See also: model accuracy by region · data sources & methodology · the data as JSON.