How the Atlas measures football.

One scale, one neutral point

Every Atlas metric publishes on 0–100 with one decimal, so a player, a club, a competition and a nation can be read side by side without conversion. 50 is the neutral point, not zero. A metric with no evidence behind it sits at 50; a score below it means the evidence is genuinely poor, not merely thin.

The six metrics

Atlas Rating Player · 0–100

On-pitch output, the provider’s own match rating, availability and the level played at, blended and then regressed toward the neutral point in proportion to minutes played.

Atlas Club Rating Club · 0–100

The quality of a squad, folded from the Atlas Ratings of its players and the strength of the competition they play in.

Atlas League Rating Competition · 0–100

The published measure of a competition, computed over its refined strength — the number a league is ranked by.

League Strength Competition · 0–100

How difficult a competition is to perform in: coverage, foreign share, nations represented, the provider’s mean rating and a curated prior. It is the input every player is rated against.

Country Index Nation · 0–100

A nation read through its players: the quality of its best (top-N mean Atlas Rating), the reach of its diaspora (players abroad, distinct leagues and clubs, and how strong those leagues are) and its depth.

Development score Nation · 0–100

The youth pipeline: the share of U21 and U23 players, the minutes those players are actually given, the age of a first export, and retention.

Thin samples are regressed, not celebrated

The failure mode of every naive per-90 ranking is the substitute who scores in his only twenty minutes and reads as the best player alive. Atlas Rating blends output, match rating, availability and level, then pulls the result back toward 50 in proportion to the minutes behind it. A player needs a real sample before he can reach a real score, which is why a squad player's rating climbs through a season rather than arriving intact in August.

Positions are scored on their own terms. A goalkeeper is never scored on goals or assists: shot-stopping and goals conceded carry his rating instead, and conceding is discounted by the difficulty of the competition, because it is a team outcome before it is a personal one.

The circular dependency, and how it is broken

A league is strong because its players are good; its players rate highly because they play in a strong league. The engine refuses to iterate that loop — a published number would then depend on how many passes happened to run. Instead it runs a fixed, acyclic pipeline, once, in this order:

1. Seed League Strength from provider facts only — coverage, foreign share, nations represented, the provider's mean rating, and a curated prior. No Atlas Rating exists at this point, by construction.
2. Atlas Rating for every player, measured against that seed.
3. Atlas Club Rating for every club, from those player ratings and the seed.
4. Refine League Strength: one blend of the seed with observed quality.
5. Atlas League Rating, published over the refined strength.

Players are always rated against the seed. The refined strength is a presentation number and is never fed back. The trade is one pass of accuracy for a pipeline that returns the same answer every time it runs, on the same data.

Percentiles are published with their population

A percentile means nothing without the group it was taken against, so the Atlas never prints one on its own. Every percentile on a player profile is computed inside a named cohort — a position group, a band of league strength and a minutes floor, for example midfielders in top-five leagues with 900+ minutes — and the cohort, including how many players are in it, is stated beside the bar. Where too few players qualify, the band widens and then the floor drops, and the published label changes with it.

Change is shown with direction

Movement is never a bare number. Rating changes carry a sign and a true minus — +3.2, −0.8 — and rank movement is written as ▲ 4, ▼ 2 or an em dash for no change. A climb from eighth to fourth is ▲ 4: the direction is the point, and it is a distinct type in the engine for that reason.

What this covers, and what it does not

The published season is 2025/26. It covers 35 competitions, 616 clubs and 22,772 players from 151 nations. A nation with no tracked players has no Country Index, and the Atlas says so rather than scoring it.

The coefficients behind every metric live in one calibrated weight set, not scattered through the code, and they are expected to be re-fitted once a full season has been ingested. Until then the scores are internally consistent and comparable with each other, and should be read as the Atlas's ranking of the same season — not as a settled historical rating.

Where the underlying data cannot answer something the design asks for, the Atlas leaves it blank and records why. Nothing on this site is estimated to fill a gap.

The terms used across these screens are defined in the glossary, and everything the Atlas ingests is listed under data sources.