How each team plays

Summary

Is a team high-pressing or deep-defending? Do they build up patiently or play direct? This dataset captures each WC 2026 team's tactical fingerprint across 12 playing-style dimensions.

The dataset lives at data/wc2026/team_style_vectors.csv, one row per participant (48 total — including placeholder rows for teams with zero coverage). It is produced by scripts/build_team_style_vectors.py from the StatsBomb open dataset; per-match feature tables cache to data/raw/statsbomb_features/ (gitignored) so re-runs against the same fixture list are cheap. Teams with no StatsBomb coverage are backf…

Sample

What each row carries

Column	Definition
`possession_share`	Fraction of total possessions the team had.
`ppda`	Passes per defensive action — a press-intensity proxy (lower = higher press).
`directness_index`	A measure of how directly the team moves the ball goalward; higher = more direct.
`build_up_passes_per_attack`	Average number of build-up passes per attacking sequence.
`crossing_rate`	Crosses per 90.
`long_ball_rate`	Long passes per 90.
`set_piece_routine_variety`	Distinct corner / FK routine signatures observed.
`final_third_entries_per_90`	Entries into the attacking third per 90.
`shots_per_90`	Self-explanatory.
`xg_per_shot`	Mean xG quality per shot.
`matches_used`, `tournaments_used`	Sample size, plus the tournament labels the rows came from.
`data_freshness`	Most recent match date used.
`style_source`	Which source the row's style read comes from: `statsbomb` (≥3 StatsBomb matches), `statsbomb_thin` (1–2), `fotmob` (no StatsBomb coverage — FotMob season stats used), or `none`.

…

Full document

Pro Pass

Want the full document?

How each team plays runs 961 words. The Pass unlocks this document and every research note in full, plus per-fixture probabilities, the four-model comparison, and per-fixture tactical analysis.

View pricing →

24h self-service refund·No subscription, no auto-renewal·Access through 31 Dec 2026. See refund policy.