Football is turned into numbers in two very different ways. Event data records the discrete on-ball actions in a match — every pass, shot, tackle and touch — tagged with who, where and when. Tracking data records the continuous position of all twenty-two players and the ball, captured many times a second. Almost every statistic you read comes from one of these two sources, and knowing which one changes how much you should trust it.
Event data is a log of things that happen to the ball. Each on-ball action is recorded as an entry with a type, a pitch location, a timestamp and a player. String those entries together and you get a structured account of the match: a sequence of passes, carries, shots and duels that can be counted, mapped and modelled. Historically this logging was done by trained human analysts watching the game; it is now increasingly assisted or automated, but the output is the same shape — a list of discrete events.
Tracking data is different in kind. Instead of logging events, it samples positions. Cameras or wearable devices capture the x-y coordinates of every player and the ball at a high frequency, roughly twenty-five times per second. The output is not a list of actions but a continuous map of movement, in which every run, shuffle and drift is recorded whether or not the player is anywhere near the ball. One source tells you what was done. The other tells you where everyone was while it was being done.
The clearest difference is coverage of the pitch. Event data is on-ball only. It captures the player in possession and the action they take, but it is silent about the other twenty-one players in that instant. The striker peeling into space, the full-back holding a defensive line, the midfielder screening a passing lane — none of them appears in event data unless they touch the ball.
Tracking data captures all of it. Because it samples every player continuously, it sees off-ball movement as clearly as on-ball action. The decoy run that drags a defender away, the shape a defence holds without the ball, the gaps that open and close as play develops — these are invisible to event data and native to tracking data. This single distinction drives almost every other difference between the two.
Because they capture different things, they answer different questions. Event data is built to answer "what happened and who did it." How many progressive passes did a midfielder play? Where did the shots come from, and what was their quality? Metrics like expected goals are event-based: they take the recorded context of a shot and estimate its chance of scoring. Expected threat and possession-value models work the same way, assigning value to on-ball actions that move the ball into more dangerous areas.
Tracking data is built to answer "where was everyone, and how did space move." How compact did a defence stay? How far did the pressing line push up? How much of the pitch did a team effectively control at a given moment? Concepts such as pitch control and off-ball run value can only be built on positional data, because they depend on where players are when they do not have the ball. If your question is about structure, space or physical output, event data cannot reach it.
This is where the practical divide appears. Event data is comparatively cheap to produce, is collected across a very large number of competitions, and is broadly standardized, which makes it the workhorse of public football statistics. When a website shows you passing numbers, shot maps or expected goals for dozens of leagues, it is almost certainly event data underneath.
Tracking data is expensive. It requires camera rigs installed at grounds or wearable systems, heavy processing, and storage of enormous volumes of coordinates. As a result it is often limited to top divisions, frequently held privately by clubs and data suppliers, and less consistent from source to source. That is why deep off-ball analysis remains largely the domain of professional recruitment and coaching departments, while fans, bettors and fantasy players work mostly with event-derived numbers. The gap is narrowing, but it is real.
Neither source is raw truth. Event data depends on consistent logging, and definitions can differ between providers — one supplier's "key pass" or "tackle" may not match another's, so totals are not always comparable across sources. It is also blind to context by design, since it records the action and not the surrounding picture.
Tracking data is objective about position but demands enormous processing before it means anything. Coordinates must be calibrated, cleaned and turned into models to become insight. And that is the key point people miss: both sources rely on a layer of modelling to produce the metrics we actually use. Expected goals is a model sitting on top of event data. Pitch control is a model sitting on top of tracking data. The number is only ever as good as the model above the raw feed, whichever feed it is.
For most of their history the two feeds were produced in completely separate ways. Event data came from analysts logging actions by hand, a craft refined over decades into a fast, structured stream. Tracking data arrived later, first through camera systems installed at grounds and through wearable devices used in training to measure physical load. The two lived in different worlds: one a manual record of actions, the other an automated record of positions.
That separation is now breaking down. Computer vision has advanced to the point where software can watch match footage and derive event data directly from it, detecting passes, shots and duels automatically from the same video that produces the tracking feed. In other words, the positional stream is increasingly being used to generate the event stream, rather than the two being collected independently.
This convergence has real consequences. It lowers the cost of event data, because less of it needs manual logging. It raises consistency, because automated detection applies the same rules every time rather than relying on human judgement. And it opens the door to metrics that fuse both layers natively, since they were captured from the same source in the first place. For now the practical divide still holds — broad, affordable, cross-league numbers come from event data, while deep off-ball detail comes from tracking — but the clean line between them is fading, and knowing that helps you read where the game's data is heading.
The two sources are not rivals so much as complements, and the frontier of analysis is combining them. Synchronizing an event stream with a positional stream lets an analyst ask far richer questions: not just whether a pass was completed, but how much space the receiver had and how many defenders it took out of the game. Possession-value models grow sharper when they know both the action and the positions around it.
For most public purposes, event data does the heavy lifting on its own, because it is available, comparable and enough to answer the majority of questions fans ask. Tracking data adds the off-ball dimension when the question demands it. A platform such as RubiScore presents event-derived figures — shots, passing, expected goals and the rest — in a form built for reading a match rather than running a recruitment model, which is what most viewers actually need.
A simple rule of thumb covers most cases:
Tracking data versus event data is not a contest with a winner. They capture different halves of the same game — one the actions, the other the positions — and the right choice depends entirely on the question. For the public game, event-derived data is the reliable, comparable workhorse that powers nearly everything you read. Tracking data is the specialist instrument that reveals what happens away from the ball.
The habit worth building is simply to ask where a statistic came from. A number drawn from event data and a number drawn from tracking data are answering different questions, and treating them as interchangeable is how analysis goes wrong. Event-based match statistics for teams, players and fixtures are published on rubiscore.com.