How our odds work

The one sentence we quote

Every bet on this site is an instance of a single sentence:

"Delivered 12″+ in 7 of the last 10 seasons — 2 were busts."

That's it. We look at a resort, a window of the winter (say, Jan 1–15), and a snowfall threshold, and we count how many of the last 10 seasons actually delivered that much snow in that window. If a future feature can't speak its claim in that sentence, we change the feature — not the sentence.

Where the numbers come from

The snowfall history comes from ERA5, a scientific weather reconstruction of the last 40+ years, read at each resort's summit. Reconstructions are great at patterns — which winters were big, which windows deliver — but they run low on raw amounts. So we anchor each resort's magnitude to its industry-standard reported average snowfall: one calibration number per resort per year, applied to everything equally. That anchor is the same number skiers already calibrate their expectations against.

Honest caveat: marketed snowfall averages are generous, and some are more generous than others. We bound that two ways — by cross-checking against mountain snow stations (SNOTEL) where they exist, and by flagging any resort whose calibration looks extreme next to its terrain peers for manual review. And each summer, the grading loop below retunes the anchor wherever we systematically over- or under-delivered against our own quotes.

The threshold: your week, your foot of snow

The default bet is 12″ of new snow per 7 days — a foot for your week. Longer windows scale it up proportionally, rounded to clean numbers (a 15-day window bets 24″; a full deep-winter month bets 54″). You can also pick 6″ (a refresh), 18″ (proper), or 24″ (deep) per week. The threshold means the same thing everywhere — what changes between mountains is the count, which is exactly where the difference belongs: Alta might deliver a foot-a-week in 9 of 10 seasons; an eastern hill in 3 of 10. Both numbers are true, and both are useful.

The labels

The label is a fixed, published mapping from the last-10 count. Nothing else moves it:

8–10 of 10	Strong bet
6–7 of 10	Solid odds
4–5 of 10	Coin flip
0–3 of 10	Long shot

If fewer than 8 usable seasons exist for a window, we show you the seasons themselves and no label — data without a verdict. Deeper views also show the last 20 seasons ("14 of the last 20"), and when the recent 10 run meaningfully below the longer record, a "softened lately" note says so out loud.

Busts — named before the bet

A bust is a season that came in under one-third of the bet — for a 12″ bet, under 4″. That's the trip that stings, so it's counted and shown before you commit, not after. One caveat we put everywhere: our odds count natural snowfall only. At snowmaking-dependent resorts, a low-natural-snow week can still ski fine on man-made cord — the odds describe snow from the sky, never your whole trip.

Odds are not a forecast

Weather forecasts see about 16 days out. Beyond that line, anyone quoting you a number for your February trip is doing climatology, whether they admit it or not — we admit it. Inside 16 days, we show the forecast next to the original odds, clearly labeled; the two are never blended into one mongrel number. A quote made in September stays visible in February, exactly as it was made.

How we grade ourselves

Every quoted odd is logged the moment it's made — resort, window, threshold, count, label, and the methodology version it was quoted under. The log is append-only: rows are never edited, re-graded, or deleted, ever. When the season's final data arrives each summer, every window is graded hit (made the bet), miss (under it), or bust (under a third of it), and the season report is published in the same counting language: "Strong bets: quoted 214 times, delivered in 83% of them."

We grade ourselves systematically, not just where traffic landed. A scheduled job — the house book — quotes every resort, every window, every threshold on our standard grid at the start of each season: thousands of standing quotes, including mountains nobody browsed that week. House quotes are marked as such forever and reported separately from quotes real visitors saw. Framed plainly: we can't cherry-pick our wins, because we bet the whole board.

We also grade during the winter, not just at the end of it — those grades are the ones marked provisional, and the next section explains exactly what that word is doing.

What "provisional" means

Grades on the scoreboard arrive in two stages. The early ones are labeled provisional, and here's exactly what that word is doing.

About a week after a betting window closes, we check what actually fell and mark the bet hit, miss, or bust. That goes on the scoreboard right away. The catch: snow data that fresh is still preliminary — the science agencies keep revising it for another two or three months. So an early grade is a real reading, honestly made, that can still move.

The official grade comes the following summer, once, from the finalized data. That one is the record. A provisional grade never overwrites it, never changes the odds we quoted you, and never quietly disappears if it turns out to be wrong.

What	Provisional	Official
When	~1 week after the window closes	The following summer
Data	Still being revised	Final
What it is	An early read	The record
Can it change?	Yes	No

One rule worth naming: if any day inside a window is missing from the data, we don't grade that window at all — we wait for the rest and try again the following week. A grade with a hole in it isn't a grade.

Why do it twice? Because the summer grade is the one that counts, and we'd rather not run that process for the first time on the season it's judging. Grading every week in public means we've practiced all winter where you can see it — including if it goes wrong.

The record itself lives on the scoreboard, which is public and stays public whether it flatters us or not.

There's one more number we intend to publish: how often the early call matched the final one. We don't have it yet — no season has finished under this system. When we do, it goes on this page whatever it says. Until then, take this section as a description of what we do, not a claim about how well we do it. That's a claim we haven't earned yet.

Ranking integrity

The odds engine is a sealed box: resort, window, threshold in — counts out. It has no knowledge of partners, payments, memberships, or anything else about you. That's enforced structurally: an automated test in our build fails if the odds code so much as imports from partner or monetization code. Partner status never enters the ranking, and any partnership that exists is disclosed where it appears.

The El Niño lens

El Niño and La Niña genuinely tilt some mountains' odds, so we offer them as an exploratory lens — "in 13 El Niño winters since 1985, this window delivered 12″+ in 8." The lens never changes a headline number, and small samples show seasons, not verdicts.

Why medians, not averages

Snowfall is streaky: a few monster seasons drag averages up past what a typical winter feels like. So "typical" on this site is the median — the middle season — with the middle-half range around it. Early versions of this project used averages and modeled scores; we retired them because honest and repeatable beats sophisticated and opaque.

What we never do

• No decimal probabilities — 10–20 seasons of history cannot honestly support "73.4%".
• No blending forecast into odds, or odds into forecast.
• No labels on samples smaller than 8 seasons.
• No inches-per-dollar rankings.
• No long-term trend claims from reanalysis data — measurement methods drifted across decades, and we won't dress that drift up as climate signal.
• No "full coverage by ⟨date⟩" promises — terrain coverage is an operational fact we can't measure honestly from here.
• No paid placement in the odds, ever. No undisclosed partnerships, ever.
• No edits to the grading log. The record stands, warts shown.

Versioning

The method itself is versioned — every quote records the version it was made under (currently v1.2.0), the math is frozen during the season, and any change lands at a season boundary with a public note here. Old quotes are never re-graded under new rules.