Player types · families

The Archetypes

Thirty player types, grouped into eight cross-position families — the Silva mould, which profiles a player by his most natural tendencies rather than the badge on his shirt. A Regista and an inverted full-back can share a fingerprint; a nine and a shadow striker need not. This is the map of how footballers actually behave on the ball.

— players classified 30 archetypes 8 families Understat on-ball data Big-5, 2025/26

The taxonomy at a glance

Season

Pick a season to reload the whole map — see the Registas of 2023/24 against the Registas of 2025/26. Percentiles, roles and archetype fits are computed within each season's own cohort, so a fit always means "against that season's peers". The Statz Rating is cross-season comparable — its absolute-output benchmarks are pinned across every season.

Every outfielder and keeper who cleared the minutes bar is scored against a library of role signatures — weighted bundles of per-90 on-ball metrics, z-standardised within his own position so a centre-back is only ever compared to other centre-backs. The archetype he fits best becomes his primary; a close runner-up, his secondary. Browse the types below by family or by position, then open any card to read its definition and the full ranked list of the players who wear it.

Browse the archetypes

Browse

Eight cross-position families

How the archetypes are computed

Each archetype carries a signature: a bundle of per-90 on-ball metrics with a weight and a sign — + for "does more of this than positional peers", − for "does less". A player's metrics are z-standardised inside his position group (CM and DM share one pool, matching how the roles overlap), then the weighted signed sum gives a fit z-score for every eligible archetype. His best fit is the primary; a second archetype within Δz = 0.5 becomes the secondary. Confidence is high when the top fit clears z = 0.5, medium above 0.35, else low. A player needs at least 60% of a signature's metrics present to be scored at all. Goalkeepers are gated on a separate two-axis model — distribution against shot-stopping — rather than a signed signature.

This is honest about what it can and cannot see. Every input is shape-inferred from on-ball event data: shots, passes, carries, key passes, tackles, interceptions, expected-goal chains. There is no tracking data, no aerial-duel or hold-up record, no defensive-positioning feed — so a physical target man and a small fox in the box can look identical on the ball, and a brave line-holder cannot be separated from a deep-sitting sweeper by height alone. Read the caveat on every type; it names the specific blind spot.

The Archetypes

01 The taxonomy at a glance

02 Browse the archetypes

The taxonomy at a glance

Browse the archetypes