Sports Taxonomy Engine
Domain-specific taxonomies for every major sport
Every event tree below is a working schema, not a demo. Annotators tag hierarchical event groups, sub-events and qualifiers against real pitch, court and field coordinates — so your model learns the rules, not just the pixels. Pick a sport, then select any row to see the exact record it produces.
Five layers of depth
Each annotation resolves to a full path. Selecting a row below lights up the path it fills.
| Event group | Sub-events | Qualifiers / metadata |
|---|---|---|
| 01Game State | First/Second Half (Start/End), Extra Time, Kick-Off, Delay (Start/End), Ball In/Out of Play | Home/Away, Attacking Direction (L/R), Team Formation |
| 02Passes | Basic, Back, Build-Up, Long, Diagonal, Throw-In, Cross, Deep Cross, Cut Back, Head, Through, Key Pass, Switch | No/Medium/High Pressure, Successful/Unsuccessful, Pass Length (m), Progressive |
| 03Reception | Successful Reception, Missed Reception, First Touch, Aerial Reception | No/Medium/High Pressure, Body Part |
| 04Possession Action | Ball Possession, Save Possession, Loss Possession, Dribble, Touch, Shield, Carry | Dribble Successful/Unsuccessful, Take-On vs 1 / 2+ Defenders |
| 05Defensive Action | Block Pass, Block Shot, Interception, Tackling, Sliding Tackle, Clearance, Dribble Against, Recovery | Successful/Unsuccessful, Defensive/Middle/Attacking Third |
| 06Pressing & Off-Ball | Press Trigger, Counter-Press, Pressing Trap, Defensive Line Height, Off-Ball Run (Overlap/Underlap/Third-Man/In-Behind) | Run Received/Ignored, Pressure Duration (s) |
| 07Duel / Attacking Action | Ground Duel, Aerial Duel, Counter Attack, Progressive Carry, Fast Break | Win/Lose, Successful/Unsuccessful |
| 08Goalkeeper Actions | Save (Low/High), Goal Kick, Punch, Catch, Claim, Sweeper Action, Distribution (Throw/Kick), Leaving Line | Successful/Unsuccessful, Shot-Stopping Difficulty |
| 09Set Pieces | Corner (Inswing/Outswing/Short), Direct & Indirect Free Kick, Throw-In (Long/Short/Foul), Drop Ball | First/Second Contact, Delivery Zone |
| 10Corners | Far Post, Near Post, Medium, Top-of-the-Arc, Short, Skip Header, Direct Score | Chance Created, Ineffective |
| 11Penalties | Awarded, Scored, Saved, Missed, Retaken | Placement (9-zone), Keeper Dive Direction |
| 12Fouls / Cards | Regular, Hand Ball, Violent, Simulation, Protest; Yellow (First/Second), Red Card | Free Kick Awarded, Advantage Played, Time Lost |
| 13Offside | Offside Called, Offside Trap Sprung, Marginal Call (VAR) | Attacking/Defending Side, Frame-Accurate Line |
| 14Free Kick / Shot | Shot, Pass; On Target, Off Target, Blocked, Fanned, Woodwork | In Game, After Free Kick, After Penalty, After Corner |
| 15Shot Dir / Foot | Bottom/Middle/Top × Left/Centre/Right (9-zone goal grid) | Right Foot/Left Foot/Head/Body, Volley/Half-Volley |
| 16Goals / Assists | Goal, Goal Conceded, Own Goal, Disallowed; First/Second Assist | After Shot/Penalty/Ricochet, Big Chance |
| 17Substitutions | Tactical/Injury Substitute, Concussion Sub, Injury (Start/End), Positional Change | In/Out, Minute, New Position |
| 18Player & Ball Tracking | Player XY, Ball XY(Z), Speed, Distance, Sprint Count, Team Shape Centroid | Pitch-Normalized 105 × 68 m, Frame Timestamp |
Need one of these schemas mapped to your own footage?
Bring a clip and your label spec. We'll walk through the event tree your model actually needs.
The layer every sport shares
Sport-specific trees sit on one common record spine. It is what makes cross-sport training sets joinable, auditable and safe to ship.
Timing
- Frame ID and wall clock
- Game clock to 1/1000 s
- Period and phase context
- Event duration (start/end)
Identity
- Player, team and jersey
- Position and lineup slot
- Officials and substitutions
- Stable IDs across matches
Space
- Normalized (x, y) — (z) where tracked
- Declared origin and axis direction
- Zone label alongside raw point
- Camera and homography reference
Quality
- Annotator ID and pass number
- Inter-annotator agreement
- Per-field confidence
- Review state and audit trail
Sample schema shown. Full proprietary taxonomies extend to five layers of depth across 20+ sports, with per-league qualifier sets and custom event groups on request.
Next step
Build your taxonomy
with our analysts
The schemas above are the starting point. On a 30-minute call we map them to your sport, your league and your model's failure cases — then annotate a sample so you can judge the quality yourself.
- A walkthrough of the event tree for your sport
- Qualifier sets mapped to your label spec
- A free annotated sample from your own footage
- Pricing against your actual frame volume
No commitment · Sample back within 48 hours