Accuracy
Forecasts served on the site
The models behind the lift and piste ranges on the site now, from the current forecast publication, and how often each model's range held when it was back-tested. Aggregate figures over all resorts: no per-resort or per-season figures.
Publication 1203 · as of 24 October 2026
Lifts open, per week
Each range is the central 80% of the modelled distribution.
Each range is the central 80% of the modelled distribution.
| Month | Range held |
|---|---|
| November | 4,948 of 5,701 · 87% |
| December | 25,132 of 30,324 · 83% |
| January | 39,649 of 46,689 · 85% |
| February | 33,633 of 39,809 · 84% |
| March | 28,530 of 33,055 · 86% |
| April | 7,227 of 8,547 · 85% |
Same week last year, as a range of the same width: 15,420 of 16,128 · 96%
Target met: at least 75%, and more often than the same week last year.
A range spans 42% of the total on average. A wider range holds more often, so this does not make the forecast validated.
Held-out: 92 resorts never used to build or tune the model, scored on the 2025/26 season it had not seen.
Each range is the central 80% of the modelled distribution.
Each range is the central 80% of the modelled distribution.
Same week last year, as a range of the same width: 13,588 of 14,132 · 96%
Target not met.
A range spans 56% of the total on average. A wider range holds more often, so this does not make the forecast validated.
Held-out: each resort scored without its own history, on the 2025/26 season.
Pistes open, per week
Each range is the central 80% of the modelled distribution.
| Month | Range held |
|---|---|
| November | 5,044 of 5,701 · 88% |
| December | 25,333 of 30,324 · 84% |
| January | 39,477 of 46,689 · 85% |
| February | 33,914 of 39,809 · 85% |
| March | 28,762 of 33,055 · 87% |
| April | 7,392 of 8,547 · 86% |
Same week last year, as a range of the same width: 14,155 of 15,645 · 90%
Target met: at least 75%, and more often than the same week last year.
A range spans 46% of the total on average. A wider range holds more often, so this does not make the forecast validated.
Held-out: 89 resorts never used to build or tune the model, scored on the 2025/26 season it had not seen.
Each range is the central 80% of the modelled distribution.
Each range is the central 80% of the modelled distribution.
Same week last year, as a range of the same width: 12,130 of 13,446 · 90%
Target not met.
A range spans 58% of the total on average. A wider range holds more often, so this does not make the forecast validated.
Held-out: each resort scored without its own history, on the 2025/26 season.
Forecasts are estimates, not guarantees — on the day, always check the resort’s own board before you ski.
Current models
Every forecast model in the Open Lifts model registry and whether any forecast is being served on the site. Taken from the registry and the live serving state, not written by hand.
Per evaluation cell: rolling2024_seen/near FAIL · rolling2024_seen/long FAIL · rolling2024_fresh_resorts/near FAIL · rolling2024_fresh_resorts/long FAIL · fresh_resorts_2025/near PASS · fresh_resorts_2025/long FAIL · viewed2025_seen/near PASS
Per evaluation cell: rolling2024_seen/near FAIL · rolling2024_seen/long FAIL · rolling2024_fresh_resorts/near FAIL · rolling2024_fresh_resorts/long FAIL · fresh_resorts_2025/near PASS · fresh_resorts_2025/long FAIL · viewed2025_seen/near PASS
Lifts open, daily (reconstruction)
The real, current evaluation of the Open Lifts open-lift-count model — every split, the baseline it's measured against, and exactly how much evidence it's based on. This page is generated directly from the same evaluation artifacts the model itself was registered with — nothing here is written by hand.
Model reconstruction-hgb-v1-20260918 is currently labelled EXPERIMENTAL. The PRD requires ≥75% range hit rate, beating a plain baseline, on every held-out split below before a model may be called validated. This one has not cleared that bar yet — every prediction it produces is experimental, not a booking-grade promise.
model held-out-date coverage 0.743 does not beat the resort-historical-mean baseline's 0.788 — PRD §2.9.1 requires both; held-out-date coverage 0.7428571428571429 < 0.75; held-out-resort evidence insufficient: only 3 eligible resorts (need >=10 per Phase 3E Sec 14's own criterion for a trustworthy estimate); held-out-season evidence insufficient: only 79 training rows (Phase 3E Sec 14 flags 300+ as the scale needed to trust this split); held-out-season coverage 0.5263157894736842 < 0.75
The three required splits
A range "hit" means the actual observed open-lift count fell inside the predicted range. Each split answers a different question about whether the model actually generalises, not just whether it fits data it has already seen.
The baseline comparison
A baseline is the simplest reasonable guess with no model at all — here, a resort's own historical average for that time of year. A real model has to beat it, not just clear 75%, or it isn't adding anything.
This model currently does not beat the baseline on this split — the honest reading is that a simple historical average is at least as good a guess right now.
Why the gate isn't just a percentage
A coverage number on too little evidence isn't trustworthy on its own. Here's exactly how much real evidence this evaluation is based on:
Live forecast verification
Separately from the training-time backtest above, every real forecast this model makes is checked against what actually happened once the date passes — the ongoing, real accuracy record.
Model reconstruction-hgb-v1-20260918, trained on data up to 16 June 2026, evaluated 18 September 2026. See how it works for what observed/reconstructed/forecast mean, or /data for source attribution.