How PassCast works

PassCast answers one question: is a 2026-27 season pass worth it for how you actually ski? It does that by simulating the coming winter 10,000 times per mountain, every day, and running each simulated winter through breakeven math against real pass prices.

1. The analog engine (the core)

For each of 53 mountains across 11 countries we hold 36 real seasons (1990 through 2026) of daily ERA5 reanalysis snowfall at mid-mountain elevation — Nov-Apr winters in the north, Jun-Sep seasons in the south — each tagged with its seasonal Oceanic Niño Index (DJF for northern winters, JJA for southern ones). ERA5's ~25 km grid systematically underestimates ridge-top snowfall, so each mountain's record is bias-corrected to its resort-reported long-term average — the year-to-year variability and ENSO response come from physics-based reanalysis, the absolute level from the mountain's own record.

Winter 2026-27 is a near-lock El Niño (NOAA CPC, June 2026: 99% for DJF, 63% chance it peaks very strong) — and the same event is shaping the southern season underway right now. The simulation resamples each mountain's historical seasons with weights set by those live phase probabilities and by closeness to the expected event intensity (Niño3.4 ≈ +1.9°C) — so strong El Niño analogs like 1997-98, 2015-16, and 2023-24 dominate the pool, at that specific mountain. Sampling noise is added because 36 seasons is a finite record. Where ENSO genuinely has little say — the Alps, whose winters ride on the unpredictable North Atlantic Oscillation — the conditioning is correspondingly weak and the page says so rather than manufacturing a signal.

2. Live seasonal tilts (updated daily)

• NOAA CPC seasonal outlook — we point-query CPC's DJF precipitation outlook polygons for each US mountain daily. A wet or dry lean tilts simulated totals, but shrunk ~50%, because CPC's winter lean is itself mostly ENSO — which the analogs already carry. Double-counting a driver is the classic way seasonal forecasts oversell themselves. Outside CPC's coverage, a literature-based regional ENSO tilt (Masiokas for the Andes, EAWM research for Japan, BoM for Australia, NIWA for NZ) fills in at partial weight.
• ECMWF SEAS5 51-member ensemble — the actual dynamical seasonal model, fetched daily per mountain out to ~7 months. We compare its ensemble-mean winter snowfall against the same grid point's ERA5 normal and apply the lean shrunk 75% and clamped to ±10%. Labeled experimental: seasonal models and reanalysis carry different biases, so we treat it as a nudge, never a level.
• Live ENSO observations — the latest ONI value is fetched from CPC daily; phase probabilities update with each monthly CPC/IRI advisory.

3. In-season blending (live right now in the south)

Once a season is within reach of real forecasts, simulation hands weight over to observation: snowfall already on the ground (ERA5 season-to-date) plus a blended 16-day multi-model forecast (GFS, ECMWF IFS, ICON, GEM) become known, and only the remaining fraction of the season is simulated. For northern mountains in July that known fraction is 0% — and the driver panel says so rather than pretending a 16-day model run matters for February. For Chile, New Zealand, and Australia the 2026 season is underway, so their PassCasts are running this blend live today — the pulsing markers on the map.

4. The skiability rating

Percent-of-average alone can't rank skiing: a marginal hill at 150% of a 55" typical season still skis worse than Alta at 90% of 494". The 0-100 rating blends four things, each shown with its own bar on every mountain page:

• Snow quantity (35%) — the simulated median season in absolute inches, on a saturating curve (~55 points at 200", ~80 at 400"). Depth is most of what makes skiing good.
• Season lean (30%) — this season vs the mountain's own typical winter; the year-specific signal from ENSO analogs, outlook tilts, and live observations.
• Powder days (20%) — the median count of 6"+ days in the simulation; 25 such days scores full marks.
• Bust risk (15%) — the probability the season clears 70% of typical. A mountain that almost never has a truly bad year earns points a boom-or-bust one doesn't.

Not modeled (yet, and the pages say so where it matters): snowmaking, terrain quality, crowding, and rain-vs-snow at marginal elevations — reanalysis counts snowfall, not settled base, so warm-storm regions can post high "snowfall" that skis thinner than the number suggests.

5. From snowfall to a verdict

• Each simulated winter adjusts how much you'd actually ski: people cut days hard in lean winters and add a few in deep ones (elasticity between 0.45× and 1.15× your plan).
• Each pass is valued at what those days would have cost at the window, honoring day caps — Ikon's 7-day partner limits (including shared allotments like SkiBig3 and NZSki), Mountain Collective's 2 days + 50% off, Epic Local's shared 10 days at Vail/Beaver Creek/Whistler, Epic's consecutive-day European blocks. Multi-mountain plans are valued across every leg jointly, with simulation draws paired so a good winter is good everywhere — conservative for geographically hedged trips.
• Shared allotments are allocated, not duplicated: a plan with Banff Sunshine and Lake Louise draws from one SkiBig3 pool, spent greedily on the priciest tickets — the way a rational skier would.
• The verdict is the probability the pass beats window tickets across all 10,000 winters: ≥85% (with margin) = STRONG BUY, ≥65% = BUY, ≥45% = LEAN BUY, ≥25% = WAIT (prices step up on published deadlines — waiting has a known cost), below = SKIP.

Data sources

• ERA5 reanalysis, multi-model & SEAS5 seasonal forecasts — Open-Meteo (non-commercial use)
• ONI, ENSO advisory & probabilities, seasonal outlooks — NOAA Climate Prediction Center
• ENSO-region snowfall relationships — NOAA climate.gov analyses & bestsnow.net correlation work
• Pass prices & access rules — verified against official pass sites, July 2026
• Day-ticket benchmarks — 2025-26 peak-window actuals (2026-27 rates publish in fall; override in the wizard)

What PassCast is not

Not a promise of powder, not affiliated with any pass, and not a substitute for the two questions no model can answer: do you love the mountain enough to go even in a lean year, and is the refund/deferral policy acceptable to you? Verdicts quantify the odds; you bring the judgment.