TournamateLABS
OPEN-SOURCE 3D DIGITAL TWIN

Visualizing the Unseen 98% of the Match in Native 3D.

A browser-based WebGL engine built to ingest optical in-stadium tracking data (TRACAB, DFL, Metrica, SkillCorner) and render spatial tactics at 60+ FPS without broadcast video dependencies.

NATIVE INGESTION:ChyronHego TRACABDFL SportecMetrica SportsSkillCorner BroadcastKloppy Standard
TACTICAL ANALYSIS MODE|NATIVE THREE.JS 3D
10
M. SilvaCAM
HOME TEAM (RED)
LIVE SPEED
9.2 km/h
Jogging
DISTANCE
14.2 m
Workload
BIOMETRIC PROFILE
Height: 1.78mBuild: Athletic
00:04.217/00:14.960
FRAME 106 / 37560 FPS PRE-BAKED
●Spatial Telemetry Engine • Interactive 3D Tactical Vector Breakdown
Launch Live WebGL Engine
THE ANALYTICS GAP

Numbers Don't Win Over Dressing Rooms.
Visual Verification Does.

Modern football produces millions of (x, y, z) data points per match. Yet performance staff are forced to translate these into static 2D bird's-eye plots or manually scrub through broadcast video that only tracks the player on the ball.

TRADITIONAL APPROACHFlat Matplotlib / 2D Scatter

Static 2D Top-Down Limitations

  • Invisible Sightlines: Cannot see what the defender actually saw. Did a teammate's body occlude the passing lane? 2D circles cannot show field-of-view cone.
  • Broadcast Dependency: TV video follows the ball; 80% of off-the-ball runs and shape preservation are never captured by traditional cameras.
  • Coach Disconnect: Analysts hand managers abstract numbers like "Passing Lane Openness: 0.72". Managers ask: "Show me why my center-back didn't step out."
GOALTRACE 3D REVOLUTION60+ FPS Digital Twin

Real-Time Spatial Verification

  • First-Person Eye-Level Sightlines: Step directly into the player's boots to evaluate sightlines, head scans, and reaction delays under real pitch geometry.
  • Full Pitch Omnipresence: Reconstruct all 22 players continuously, revealing the hidden weak-side defensive line compression and blind-side blindspots.
  • Counterfactual Ghosting: Project an AI-calculated optimal avatar alongside the player to instantly communicate positional mistakes during post-match reviews.
ENGINEERING CAPABILITIES

Built for High-Frequency Spatial Precision

Four proprietary modules powering the GoalTrace WebGL pipeline from raw tracking coordinates to high-fidelity tactical inspection.

01

Kinematic Spline Interpolation

Raw tracking frames recorded at 10 Hz or 25 Hz suffer from jagged stepping. GoalTrace applies Catmull-Rom and piecewise cubic Hermite splines to bridge coordinates into silky-smooth 60+ FPS playback with velocity-adjusted procedural gait and body lean.

Sampling: 10Hz/25Hz → 60HzCubic Hermite C1 Continuous
02

Tactical Ghosting Layer

Overlay holographic counterfactual avatars alongside actual player tracking. Feed tactical models (e.g. ideal compact back four or zonal shifting) into the engine to render where a player *should* have been, displaying real-time positional error vectors in meters.

Counterfactual Delta:Δ Spatial Vector + Error HUD
03

Broadcast-Derived Inpainting

Broadcast tracking data (like SkillCorner) suffers from frequent occlusions and players dropping off-camera. GoalTrace features physics-informed multi-agent trajectory completion to extrapolate missing players until optical re-identification occurs.

Occlusion Handling:Continuous Multi-Agent Recurrent Inpainting
04

Multi-Perspective Director

Break free from fixed stadium camera gantries. Instantly toggle between TV Broadcast panning, free-orbit tactical overhead, over-the-shoulder player chase, and first-person player eye-level POV to inspect occlusions and passing lanes.

Camera Modes:Free Orbit / Chase / Player POV / 2D Top
SYSTEM ARCHITECTURE

Clean Decoupled Data Pipeline

GoalTrace isolates data parsing from 3D scene graph rendering. Preprocessing modules ingest heterogeneous tracking formats into an open, binary-efficient frame stream consumed directly by Three.js shaders.

STAGE 01 • INGESTION

Data Preprocessing

Standardized ingestion using Kloppy and OpenSTARLab parsers. Cleans raw XML, DAT, and JSON files from TRACAB, Metrica, DFL, and SkillCorner.

Output: Normalized (x, y, z, vx, vy)
STAGE 02 • PROTOCOL

Unified API Contract

Compact binary/JSON match frames with pitch coordinates, player identifiers, team kits, ball state, and tactical ghosting targets.

Low Memory Footprint • Streaming Ready
STAGE 03 • RENDERING

GoalTrace Three.js WebGL

Hardware-accelerated turf shaders, dynamic pitch control meshes, player instanced meshes, and multi-perspective camera matrices.

Native 60+ FPS • WebGL 2.0
READY FOR SPATIAL VERIFICATION?

Experience Optical Tracking in Native 3D.

Test the live engine directly in your browser with sample tracking datasets or fork the open-source repository to integrate your own club's tracking data.