How it works
Short version: Open Lifts is early. This page describes what actually exists right now, not what the product is meant to become.
What exists today
A curated catalog of resorts across the Alps — name, region, country, and the lifts, pistes and sectors each one operates. That catalog is reference data: it describes what a resort has, not what's running on any given day. You can browse it under Explore resorts.
For a small number of resorts, we also show real observed status — which lifts and pistes the resort's own bulletin reports as open, closed or on hold right now, plus real weather (recent conditions and a 14-day forecast at base, mid-mountain and summit), real holiday calendar data and real pass prices from each resort's own official source. Where the source hasn't reported something, the page says so — it never guesses, and it never shows “unknown” as closed.
We also now have a real, working open-lift-count model (see a resort's Week tab) — but it hasn't yet cleared our own accuracy bar (below), so every number it produces is clearly marked experimental, not a booking-grade promise. You can also compare two resorts side by side on real lift counts, observed/forecast openness, snow, holiday context and pass prices — no overall score, just the real facts. Or plan a trip: give it your dates and up to two priorities, and it ranks every published resort with the real signal behind each position — never a hidden universal “best resort” score.
Three kinds of evidence — never blurred together
Every open-lift number on this site is one of exactly three things, always labelled as which:
Observed — a real status reported by the resort's own bulletin, right now or on a specific past day. The strongest evidence there is: nothing modelled, nothing estimated.
Reconstructed — for a past day with no observed bulletin on file, our model's estimate of what was likely open, built from that day's real recorded weather. Labelled an estimate everywhere it appears, never shown as though it were observed.
Forecast — the same model, run forward instead of back: an estimate of what will likely be open, from the most recent real weather available. Right now that means a same-day estimate only (a “nowcast”) — a genuine week-or-months-ahead forecast needs a real weather-forecast feed we aren't yet running into it, and until we are, this site doesn't claim to predict further out than that.
Pipeline, in order: raw provider bulletins → observed status → joined with real weather, holiday and lift/piste data → the model → a reconstructed (past) or forecast (future) estimate. Every one of those steps is real for the five resorts this site currently covers.
What doesn't exist yet
Crowding scores, and a validated, multi-day-ahead season-shape forecast. The model above only currently estimates today, from today's real weather, for the reason above.
Why the forecast is marked experimental
A forecasting product usually earns trust by publishing how accurate it is. Ours doesn't clear the bar yet: on the real held-out evidence we have so far, our model's range contains the actual lift count noticeably less often than our own ≥75% target on every split that matters (a date, a resort, and a season the model never trained on) — and on the split the model looks strongest on, a plain historical-average baseline currently does better, not just as well. Until a model actually clears that bar, every prediction on this site is labelled experimental, never presented as validated. The full real scorecard — every split, the baseline, and exactly how much evidence it's based on — is public at /accuracy, not just asserted here.
Where the data comes from
Every source this site uses, its licence, and whether it's currently cleared for public/commercial use, is listed at /data.