Live since 21 Oct 2026. 3 daily forecasts kept, latest observations 22 Oct 2026, csgd v2 3e5ff3f4. 2,469 forecasts checked, 3 cases of 6+ in.
Cross-validated estimate, 2024–26: Brier skill +0.36 (days 1–3, all regions)
Collecting. Numbers show once a group has 300 forecasts on 14 days (20 cases of 6+ in for Brier skill and the curve). So far: 2,469 forecasts on 2 days. Each night adds roughly 1,000 more.
When the where2ski forecast gave this chance of 6+ in, how often it happened
Collecting: the curve appears at 20 cases of 6+ in (3 so far). Dots show the cross-validated estimate.
Collecting: the curve appears at 20 cases of 6+ in (3 so far). Dots show the cross-validated estimate.
Collecting: the curve appears at 20 cases of 6+ in (0 so far). Dots show the cross-validated estimate.
Collecting: the curve appears at 20 cases of 6+ in (0 so far). Dots show the cross-validated estimate.
By region and lead
| Lead | Forecasts | Brier skill, 6+ in | Inside likely range target 67% | Above could reach target 10% | Interval score | Quantile skill |
|---|---|---|---|---|---|---|
| Western US | ||||||
| Days 1–3 | 2,4352 events | collectingcross-val. +0.35 | collecting | collecting | collecting | collecting |
| Days 4–6 | 00 events | collectingcross-val. +0.23 | collecting | collecting | collecting | collecting |
| Days 7–10 | 00 events | collectingcross-val. +0.08 | collecting | collecting | collecting | collecting |
| Western Canada | ||||||
| Days 1–3 | 120 events | collectingcross-val. +0.39 | collecting | collecting | collecting | collecting |
| Days 4–6 | 00 events | collectingcross-val. +0.27 | collecting | collecting | collecting | collecting |
| Days 7–10 | 00 events | collectingcross-val. +0.11 | collecting | collecting | collecting | collecting |
| US Northeast | ||||||
| Days 1–3 | 00 events | collectingcross-val. +0.45 | collecting | collecting | collecting | collecting |
| Days 4–6 | 00 events | collectingcross-val. +0.23 | collecting | collecting | collecting | collecting |
| Days 7–10 | 00 events | collectingcross-val. +0.01 | collecting | collecting | collecting | collecting |
| US Midwest | ||||||
| Days 1–3 | 00 events | collectingcross-val. +0.27 | collecting | collecting | collecting | collecting |
| Days 4–6 | 00 events | collectingcross-val. +0.17 | collecting | collecting | collecting | collecting |
| Days 7–10 | 00 events | collectingcross-val. +0.04 | collecting | collecting | collecting | collecting |
| European Alps | ||||||
| Days 1–3 | 221 events | collectingcross-val. +0.49 | collecting | collecting | collecting | collecting |
| Days 4–6 | 00 events | collectingcross-val. +0.37 | collecting | collecting | collecting | collecting |
| Days 7–10 | 00 events | collectingcross-val. +0.16 | collecting | collecting | collecting | collecting |
| Eastern Canada | ||||||
| Transferred from the US Northeast, not locally verified. | ||||||
Brackets: 90% range from resampling weeks. Brier skill is against each station's own climate. Cross-validated: 2024–26, not fitted on the seasons scored.
Last storm
None yet: a storm is a day with 6 in or more at 3 or more stations.
1,322 weekly forecasts archived since 21 Oct 2026; 0 resort-weeks verified so far (0 calendar weeks).
Collecting. Skill shows once 200 resort-weeks on 3 different weeks are verified. The first week ends 23 Oct 2026; its result lands 24 Oct 2026. The third verified week lands about 7 Nov 2026.
When the where2ski forecast gave this chance of an above-normal base, how often it was
Collecting: reliability appears with the first verified weeks.
By lead
| Lead | Live vs climatology | Live vs same as today | Cross-validated | Resort-weeks |
|---|---|---|---|---|
| Week 1 | collecting | collecting | +0.62vs same as today +0.29 | 0 |
| Week 2 | collecting | collecting | +0.55vs same as today +0.35 | 0 |
| Week 3 | collecting | collecting | +0.49vs same as today +0.35 | 0 |
| Week 4 | collecting | collecting | +0.45vs same as today +0.34 | 0 |
| Week 5 | collecting | collecting | +0.42vs same as today +0.33 | 0 |
| Week 6 | collecting | collecting | +0.39vs same as today +0.32 | 0 |
Why two numbers. The cross-validated figure (+0.49 vs climatology, +0.33 vs “same as today”) uses ERA5 weather on both sides: the model was handed the weather that happened. Live forecasts must predict it, so live skill will be lower. Both are shown, labelled.
Collecting. The first month ends 30 November 2026; its result lands in early December. Reliability appears once a month has been measured.
What will be checked
| Forecast month | Verified | Status |
|---|---|---|
| November 2026 | early Dec | Waiting |
| December 2026 | early Jan | Waiting |
| January 2027 | early Feb | Waiting |
| February 2027 | early Mar | Waiting |
| March 2027 | early Apr | Waiting |
| April 2027 | early May | Waiting |
Cross-validated estimate, by lead
| Lead | vs climatology | vs same as today |
|---|---|---|
| Month 1 | +0.40 | +0.25 |
| Month 2 | +0.30 | +0.24 |
| Month 3 | +0.24 | +0.23 |
| Month 4 | +0.21 | +0.21 |
| Month 5 | +0.18 | +0.19 |
| Month 6 | +0.16 | +0.17 |
Why this is a ceiling. It is perfect-prog: ERA5 weather on both sides. Live forecasts must predict that weather, so live skill will be lower.