machine learning · behavior only — no positions, no reputations
The eight species of footballer
We measured what all 523 regular outfielders actually did per 90 minutes — pressing, passing, sprinting, dribbling, shooting — and let a clustering algorithm sort them into species. It rediscovered football's roles on its own, and it doesn't care who's famous. Find your player:
Ball hunters ×88
Press first, ask later — elite pressure and forced-turnover volume, minimal build-up duty.
Ball-playing anchors ×82
High pass volume at high completion, line-breaking from deep; their rare shots are set-piece headers.
Box predators ×44
Top of the food chain for xG, shots and threat — they live between the posts and the penalty spot.
Touchline burners ×78
Pace, width and crosses — receive in behind and attack the byline.
Midfield metronomes ×79
Volume passers who never stop offering — the tempo-setters covering the most grass.
Dead-ball architects ×40
Corners, free kicks and whipped deliveries — the set-piece and cross-volume specialists.
Quiet engines ×57
Few touches, safe ball, high speed — the shift-runners who stretch play off the radar.
Chaos dribblers ×55
Take-ons, carries and pockets — the ones who break structure with the ball glued to their boot.
Four ways to play a World Cup
The same idea applied to the 48 teams' per-match identity.
Possession architects
High completed-pass volume, patient build-up, chances without chasing.
High-press gamblers
Enormous pressing numbers and enormous shots conceded — chaos as a strategy.
Low-block survivors
Below average in every attacking metric — organized, compact, waiting.
Vertical aggressors
Line breaks, runs in behind and shot volume — fastest route to goal wins.
Method: per-90 stat vectors (22 tracking dimensions), z-scored, k-means k=8, players with 180+ minutes. Goalkeepers are excluded — their numbers don't share axes with outfielders (their twins still show on player pages). The fortune index on the tournament page comes from the same pipeline.