Italy - Serie A

What the score means

The index shows how much injury-risk attention a club's squad warrants, on a 0–100 scale set against players from the 2025/26 season. A higher index means more attention is warranted. It is not a forecast of an event, and not a probability — an index of 75.5 is not a 75.5% chance of anything.

The club risk index

A club's risk index is the mean of its players' risk indexes on the day.

A club-level average on a 0-100 scale, and not a probability for any individual.

The rating

The rating is the band a club’s index falls in. A band runs up to and including its edge, so an index sitting exactly on an edge takes the lower band.

On watch

A club is on watch when its index sits just below the next rating, inside the published margin. Seven-day movement does not decide this.

The League Risk Index

The League Risk Index is the mean of the club risk indexes on the as-of date, with every club counting equally.

Who is scored

Goalkeepers are excluded from the model and are never scored. No goalkeeper contributes to a club’s index.

How often this changes

New figures arrive with each delivery. Deliveries come twice a week, each one carrying the season to date with a row per player per day, and the surface is republished when a delivery lands.

A scheduled reconcile republishes the newest snapshot between deliveries, so a failed publish or a missed event heals without anyone intervening. A reconcile republishes what is already there; it does not add new data.

Every surface states the date its data is as of. That date belongs to the data, never to the publication.

A day publishes only when that day’s delivered history covers the whole configured cohort for the competition. A day short of the configured club count publishes nothing, rather than publishing a partial league.

If a publish fails, the surface keeps serving the last snapshot that published, with that snapshot’s own as-of date. Nothing is interpolated and nothing is substituted.

Factors, where they appear

Where factors are shown, they are associations, not causes. A factor is never the reason a number moved.

How it is produced

Proprietary machine learning, developed in-house. We publish what the figures mean and how to read them; how they are produced is not disclosed.

How we know it works

Models are selected against data held back from their development, and re-checked as new data arrives. The performance measures and their results are proprietary and are not published.

What it is not

Not medical advice, and not a diagnosis. Not a probability. Not a statement about any named individual. Not a comparison between one league and another.

No prediction about any named athlete is published on this surface. Players appear only as counts by position and rating, across the whole league.

Why we hold some of it back

The method and its evidence are what make the rating ours. A published scale, published definitions and a published as-of date are what make it usable. We publish the second and keep the first.