Capabilities — WithoutBall

2.4M+Data Points Per MatchTarget25 fpsReal-Time TrackingTarget92%Prediction Accuracy21Skeleton Joints Per PlayerPose EstimationSeason-Scale AnalyticsAnswers Linked to FootageTactical Pattern Analysis

Platform targets

2.4M+Data points per match

21Skeleton joints per player

25 fpsReal-time pose inference

92%Prediction accuracy

Prototype

See

Computer Vision

Pose Estimation

Broadcast-ready multi-player pose inference targeting 25 fps — 17 detected keypoints per player, lifted onto our 21-joint 3D skeleton, stabilized for real match footage and built to feed every downstream model.

Prototype

See

Reconstruction

Skeleton Repair & 2D→3D Lifting

Real tracking data is imperfect: joints drop out, occlusions cut sequences, 2D detections carry noise. The foundation model repairs all of it in one forward pass — denoised skeletons with the gaps filled, lifted from flat 2D video coordinates into consistent 3D motion.

Prototype

Understand

Deep Learning

Movement Signatures

Compact movement signatures for off-ball behavior — built to compare roles, styles, and tendencies across leagues and seasons.

Prototype

Understand

Embedding Space

Scalable Player Profiling

Every player’s movement signature, placed in one shared space. Scouting becomes similarity search: “find me players who move like this one.”

Research

Understand

Re-Identification

Identity from Movement

Movement style is a signature: gait, acceleration habits, coordination timing. A research direction for re-identifying a player from movement alone — across cameras, across kits, when jersey numbers are unreadable.

Prototype

Understand

Retrieval Engine

Movement Search at Scale

Find “every time a centre-back got beaten 1v1” across entire seasons — zero manual tagging; you query with movement itself.

Prototype

Understand

Pattern Recognition

Tactical Fingerprinting

Automatically characterize playing styles, formations, and transitions without a single human annotation — compare opponents like a database query.

Prototype

Understand

Unsupervised ML

Tactical Clustering

Unsupervised discovery of defensive blocks, pressing schemes, and transitions — cluster patterns without a single hand-labeled event.

Research

Anticipate

Forecasting

Predictive Runs

Short-horizon probabilistic forecasting (2–3 s), in research: surface runs, overlaps, and pressing triggers before they happen — as probabilities, not certainties.

Research

Anticipate

Biomechanics

Injury Risk Signals

Biomechanical drift can appear weeks before clinical symptoms. A research program to catch it early — every athlete measured against their own baseline, with thresholds personalized per player, role, and load.

Research

Anticipate

Rehabilitation

Return-to-Play Monitoring

Rehabilitation measured against the athlete’s own pre-injury baseline: asymmetry closing week by week, coordination returning, load tolerance rebuilding — a research program for evidence-linked progress reports.

Roadmap

Anticipate

Performance Science

Match Readiness

On the roadmap: intensity, coordination, and recovery markers combined into one transparent readiness view — for training-load and match-day decisions.