What is player tracking data?

A high-frequency record of where every player, and usually the ball, was on the field or court, sampled many times per second by optical camera systems or wearable devices. Derived fields include speed, acceleration, distance covered, and sometimes the direction a player is facing.

How is tracking data different from event data?

Event data lists actions such as passes and shots with a location for each action. Tracking data records everyone's position continuously, including players far from the ball. That makes space, pressure, and off-ball movement measurable, which event data cannot express.

Can the public access player tracking data?

Only in limited form. Full feeds are generally licensed to clubs, leagues, broadcasters, and commercial partners. Public availability comes mainly from released samples and analytics competition datasets, which are narrow in scope and governed by their own usage terms. Check each release's licence before publishing derived values.

What should you check before analysing tracking data?

Normalise attacking direction and align sequences to a reference event, then validate data quality: missing frames, identity switches between players, gaps in ball position, and implausible speeds. Also confirm the sampling rate, coordinate conventions, and units for that specific source.

What Player Tracking Data Contains

Player tracking data records the position of every player, and usually the ball, many times per second, captured by optical camera systems or wearable devices. Derived fields include speed, acceleration, distance covered, and orientation. It supports analysis of space, pressure, and movement that event data cannot express, and access is usually restricted.

What the data holds

Positions over time. Coordinates for every player on the field or court, plus the ball where the system tracks it, sampled at a fixed frequency. A single match produces millions of rows.

Derived motion. Speed, acceleration, and direction of travel, computed from consecutive positions and usually smoothed, since raw differences amplify measurement noise.

Orientation. Some systems report which way a player is facing, which matters for questions about awareness and passing options.

Event alignment. Tracking is most useful joined to event data, so that a pass or shot can be located within the spatial context that produced it. Alignment relies on synchronised clocks and is a common source of error.

Physical load. Wearable systems add metrics used by sports science teams, including accelerations, decelerations, and workload measures.

How it is captured. Optical systems use fixed cameras and computer vision to identify and follow every player, which requires no equipment on the players and depends on camera coverage and occlusion handling. Wearable systems use devices carried by the player, often combining satellite positioning with inertial sensors, and are common in training environments.

League differences. Sampling frequency, coordinate origin and orientation, units, and which entities are tracked all vary. American football tracking through the league's Next Gen Stats programme uses tags on players, while basketball and soccer have predominantly used optical systems. Conventions differ enough that code written for one source rarely runs unchanged on another.

Normalising before analysis

Two steps precede almost all tracking work. Orient every sequence so that the attacking direction is consistent, since teams switch ends, and align each sequence to a reference event such as the snap, kick-off, or shot. Skipping either means comparing frames that describe different situations.

What it makes measurable

Space and positioning. Distance between players, area controlled by a team, and how compact a defensive shape is. These are the questions event data cannot answer at all, since events record what happened rather than where everyone stood.

Pressure and proximity. How close defenders were when a pass or shot occurred, which changes the interpretation of the event itself.

Off-ball movement. Runs that create space without touching the ball, which by definition generate no events.

Passing options. Which passes were available and how risky each would have been, rather than only the pass chosen.

Physical output. Distance covered, sprint counts, and high-intensity periods, used for conditioning and substitution decisions.

Model inputs. Tracking-derived features improve shot models, pass models, and defensive valuation, because they add the context the event record omits.

Limits. Tracking says where players were, not why. Intent, instruction, and tactical plan are not in the data, and inferring them requires assumptions that should be stated.

Sports science versus tactical use

The same feed serves two audiences with different needs. Conditioning staff want physical load per player across a season, while analysts want spatial context around specific events. Those imply different aggregations and retention policies, so decide which you are building for before designing storage.

Access, cost, and practicalities

Access is restricted. Full tracking feeds are generally licensed to clubs, leagues, broadcasters, and commercial partners. Public availability is limited to released samples, competition datasets, and occasional research collaborations, which is why most public work uses event data.

Competition releases. Analytics competitions have released tracking samples that are widely used for learning and research under their own terms. Those releases are narrow in scope, shaped by the competition's theme.

Storage. Millions of rows per match means terabytes across a season. Columnar formats and partitioning by match and period are close to mandatory, and storing derived features alongside raw frames avoids recomputing them repeatedly.

Processing. Most analysis operates on windows around events rather than whole matches, so an index from event to frame range saves enormous time.

Quality checks. Missing frames, identity switches between players, and ball position gaps are common. Validate continuity, check that player counts match expectations, and flag sequences with implausible speeds before analysis.

Privacy and contracts. Tracking and wearable data about identifiable individuals is personal data, and licence terms usually restrict redistribution and permitted uses. Check both before building anything that publishes derived values.

This page describes data and method and is not betting advice.

Frequently asked questions

What is player tracking data?: A high-frequency record of where every player, and usually the ball, was on the field or court, sampled many times per second by optical camera systems or wearable devices. Derived fields include speed, acceleration, distance covered, and sometimes the direction a player is facing.
How is tracking data different from event data?: Event data lists actions such as passes and shots with a location for each action. Tracking data records everyone's position continuously, including players far from the ball. That makes space, pressure, and off-ball movement measurable, which event data cannot express.
Can the public access player tracking data?: Only in limited form. Full feeds are generally licensed to clubs, leagues, broadcasters, and commercial partners. Public availability comes mainly from released samples and analytics competition datasets, which are narrow in scope and governed by their own usage terms. Check each release's licence before publishing derived values.
What should you check before analysing tracking data?: Normalise attacking direction and align sequences to a reference event, then validate data quality: missing frames, identity switches between players, gaps in ball position, and implausible speeds. Also confirm the sampling rate, coordinate conventions, and units for that specific source.