Where every number comes from

Every verdict on this site is generated from open data by a pipeline that runs each morning. Nothing is hand-written about current conditions, nothing is estimated by feel, and every value can be traced to a named source below. This page is the full recipe.

How are current conditions computed?

Each resort page combines four inputs, refreshed daily:

• Modelled snow depth and forecast - from the Open-Meteo weather API, queried at three elevations per resort: base, mid-mountain and summit. The headline depth is always the mid-mountain value, because that is where most of a ski day happens. Modelled values are labelled as modelled.
• Measured station depth - where a GeoSphere Austria weather station sits near the resort, we show its measured snow depth next to the modelled value as a reality check, named with the station and its elevation. Models drift; stations do not.
• Avalanche danger level - the official EAWS danger level for the resort's warning region, from avalanche.report and the European Avalanche Warning Services. We display the level and link the bulletin; we never paraphrase avalanche advice.
• Season status - opening and closing dates from resort announcements, reviewed by a human each autumn. Dates estimated from historical patterns are explicitly flagged as estimated until the resort confirms.

How are the history statistics computed?

The snow-history chart on each resort page is built from the Open-Meteo historical reanalysis archive at the resort's mid-mountain elevation, covering the last 15 winters. For each half-month from November to May we compute three values:

• Reliability - the share of those winters with a snow base of 30 cm or more by that date (the 15th for the first half, the last day of the month for the second). 30 cm is a conservative threshold for a properly skiable groomed piste on typical Alpine terrain.
• Median depth - the depth on that date in the middle winter, not the average, so single freak seasons cannot flatter or slander a resort.
• Snowfall days - how often it actually snows in that half-month.

The statistics are recomputed every autumn with the most recent winter added and the oldest dropped, so the numbers track the current climate rather than a nostalgic average. Where a GeoSphere station exists, its long-term record is used to sanity-check the modelled history. Snowmaking is not modelled: real-world early-season coverage at well-equipped resorts is often better than the natural-snow statistics suggest, never worse, so the numbers err on the honest side.

When is the data updated, and what happens when a source fails?

The whole site is rebuilt daily at ~06:30 Alpine time (CET/CEST). Every verdict carries the timestamp of the data behind it. If an upstream source fails on a given morning, we keep the previous day's values and mark them stale - the page then says "updated" with the older date rather than pretending to be fresh. A verdict you read here is either today's data or honestly labelled as yesterday's; there is no third state.

What are the known limitations?

• Model vs reality. Modelled snow depth at a point elevation is an approximation of a whole mountain. The measured-station cross-check narrows the gap in Austria; Italian coverage relies more heavily on the model.
• Piste conditions are more than depth. 40 cm of cold groomed snow skis better than 80 cm of rain crust. Depth and reliability are the best available proxies, not a full description.
• Season dates change. Resorts move opening dates with the snowpack. The daily rebuild picks up announced changes, but a date can be a few days stale in fast-moving early winter.

Data sources and licences

All sources are open data, used with attribution as their licences require:

Can you use our dataset?

Yes. The computed snow-history statistics (reliability percentages, median depths and snowfall days for every resort we cover) are published under CC BY 4.0 - cite SnowVerdict with a link. Journalists and researchers who want the underlying JSON files rather than the rendered pages can write to [email protected]. Derived statistics remain subject to the upstream licences listed above.