Football data: events, tracking and physical

In one lineFootball data comes in three main kinds — events, tracking and physical — supplied by a small group of companies and turned into metrics that only make sense once you know how they are defined.

Three kinds of data

Event data

Event data is a log of every on-ball action in a match: passes, shots, tackles, fouls, clearances and more. Each entry records who did it, when, where on the pitch, and what the outcome was. Stats Perform says its Opta data is collected live by trained analysts, supported by computer vision (software that reads video). Many public football statistics, including expected goals, are built on event data.

Its blind spot is everything away from the ball. It records where a pass went, but not where the other 21 players were standing.

Tracking data

Tracking data records the position of every player and the ball many times a second, for the whole match. It can come from cameras installed around a stadium, or from ordinary broadcast video processed by computer vision. With it, analysts can measure things events cannot: the height of a defensive line, the space between players, or the runs that were never rewarded with a pass.

Some products sit between the two. Hudl StatsBomb’s 360 data adds a freeze frame to each event, showing the location of every player visible in the camera shot at that moment.

Physical data

Physical data describes how much and how hard players move: total distance, distance at high speed, sprints and accelerations. It can be calculated from tracking data, or captured by small wearable devices attached to a player’s kit. The Laws of the Game allow such wearables in official matches, as part of an electronic performance and tracking system (EPTS). The competition organiser must make sure the devices are not dangerous and meet the requirements of FIFA’s quality programme for EPTS.

Who supplies the data

The companies below are among the best-known suppliers of professional data. Ownership in this industry changes, so the table below reflects what could be checked in September 2026.

A dash means no parent company could be confirmed. Clubs, leagues, media and betting companies buy this data under licence. Public websites usually show only a small part of it.

Making numbers comparable

Per 90

Players get very different amounts of playing time. A per-90 figure divides a total by the minutes played, then multiplies by 90, so everyone is measured per full match.

Example: 6 goals in 1,080 minutes is 6 ÷ 1,080 × 90 = 0.5 goals per 90.

Per-90 figures from very few minutes swing wildly. A substitute who scores once in 30 minutes shows 3.0 goals per 90. That is why analysts usually set a minimum number of minutes before ranking players.

Percentiles

A percentile ranks a player against a comparison group. A player in the 90th percentile for a metric scores higher than 90% of that group.

The group matters as much as the number. Being in the 90th percentile among centre-backs in one league is not the same as being in the 90th percentile among all outfield players in Europe. Always check who the player is being compared with.

Free data to practise on

StatsBomb Open Data is a free dataset. It is published on GitHub by Hudl and includes competitions, match lists, event data, line-ups and, for selected matches, 360 freeze frames, all as JSON files (a common text format for data).

It comes with conditions, set out in the StatsBomb Public Data User Agreement:

• It is provided for research and genuine interest in football analytics.
• Anything you publish using it must name StatsBomb as the source and carry the StatsBomb logo.
• You may not exploit the data, or analysis made from it, commercially.
• You may not pass the raw data on to others.

SkillCorner also publishes a small open dataset of broadcast tracking data on GitHub, and asks users to credit SkillCorner.

Common metrics in one line each

• Progressive pass — in Opta’s definition, a completed pass in the attacking two-thirds of the pitch that moves the ball at least 25% closer to goal. Other providers use different thresholds.
• PPDA (passes per defensive action) — the passes the opponents are allowed in a set area, divided by the pressing team’s defensive actions (such as tackles, interceptions and fouls) in that area. Lower means more intense pressing. The area varies: Opta uses everything outside the pressing team’s own defensive third, while an early version published by StatsBomb in 2014 used the 60% of the pitch furthest from the pressing team’s own goal.
• Field tilt — a team’s share of the play in the final third, compared with its opponent’s. Above 50% means it spent more of the game in attacking territory.
• xT (expected threat) — gives every zone of the pitch a value for how likely possession there is to lead to a goal within the next few actions. A pass or carry earns the difference between where it starts and where it ends. A widely used version was published by analyst Karun Singh.
• xG (expected goals) — the probability that a shot is scored, based on similar past shots. See the xG guide.

Where to go next

• Expected goals (xG), without the maths — the most widely used data metric, explained step by step.
• Pressing, explained — the tactic that PPDA tries to measure.