How soccer analytics works

Soccer is a low-scoring game where a small number of shots decide most matches, which makes traditional stats like goals and assists slow to accumulate and easy to misread over a single season. Soccer analytics developed largely to answer a specific frustration with the scoreline: a team can dominate play, create better chances, and still lose to a single deflection or a moment of individual brilliance. The field's central tools exist to separate performance that is likely to repeat from performance that got lucky or unlucky.

The questions analytics is built to answer

Clubs and analysts ask a recurring set of questions: was a match's result a fair reflection of the play, or a fluke result likely to correct itself? Which players create danger even when it doesn't show up as a goal or assist? And where on the pitch does a team actually generate and concede its danger? None of these can be answered from the scoreline or a traditional box score alone.

The data it runs on

Soccer analytics is built on two main data types, both usually sourced from specialist providers rather than produced in-house by clubs.

• Event data records every on-ball action in a match — passes, shots, tackles, dribbles, duels — tagged with location, outcome and often additional context. StatsBomb is known for this kind of manually coded, highly detailed event data, extended by its 360 product, which adds the position of every player on the pitch (not just the ball) at the moment of each event — context that changes how a pass or a shot should be judged. Opta supplies event data at broader scale across many more competitions and sports, with a long-established accuracy track record used heavily by broadcasters and betting operators as well as clubs.
• Optical tracking data, captured by multi-camera systems installed in stadiums, records the position of every player and the ball many times per second throughout a match, independent of who is on the ball. Second Spectrum, now part of Genius Sports, is a leading provider of this tracking approach, and it is what makes off-the-ball movement — a run into space, a defensive shape, a pressing trigger — measurable at all, rather than inferred from event data alone.

Measuring chance quality: expected goals

[[expected-goals|Expected goals]] (xG) estimates the probability that a given shot results in a goal, based on factors like distance, angle, and the situation it was taken from (open play, a cross, a one-on-one with the goalkeeper), using historical data from thousands of comparable shots. A team that creates more, higher-quality chances than it concedes but loses the match has usually been unlucky in the way finishing bounces on a given day; over enough matches, results tend to converge toward what the underlying shot quality predicted. xG is widely used to judge whether a result flattered or undersold a team's actual performance, and to separate a striker in a finishing slump from one whose underlying chances have genuinely dried up.

Measuring what happens before the shot

Because so much of a match happens away from the shot itself, analysts extended the same logic upstream. [[possession-value-model|Possession value models]] assign a value to every action on the ball, not just shots, estimating how much each pass, carry or tackle changed the team's likelihood of scoring or conceding soon after — crediting a defense-splitting pass even when the eventual shot misses, and a defensive interception even when it doesn't lead directly to a goal. [[progressive-passes|Progressive passes]] — passes that move the ball meaningfully closer to the opponent's goal — and [[passes-per-defensive-action|passes allowed per defensive action]], a proxy for pressing intensity (fewer opponent passes before a team wins the ball back means more aggressive pressing), are two of the more common metrics built on event data to describe a team's style and a player's contribution to it.

How the work is done in practice

A recruitment analyst typically starts with a statistical screen — players outperforming their role's typical output on metrics like progressive passes or possession value, adjusted for the league they play in, since a stat that looks elite in a weaker league often regresses when the player moves up — before video scouts verify the underlying skill on tape. In-match, coaching staffs use event and tracking data turned around within minutes to check whether a tactical adjustment is producing better chance quality, not just more possession, since possession without shot quality does not reliably produce goals.

Common mistakes and misreadings

• Treating a single match's xG total as decisive. A team can generate a high xG total from several low-probability shots and lose fair value comparisons to a team that scored from one clean chance; the model's value is clearest over many matches, not one.
• Comparing raw counting stats — goals, assists, tackles — across leagues of very different quality and style without adjustment, which routinely overrates production from weaker competitions.
• Ignoring the shooter. xG estimates an average finisher's probability of scoring a given shot; it deliberately does not credit or penalize a specific player's finishing skill, which is why a small group of elite finishers consistently outperform their xG over long careers without that being a modeling error.
• Reading a hot or cold finishing run as a permanent change in ability, when a meaningful part of it is regression to the mean that will likely correct over the following months.
• Assuming more possession automatically means better performance. Possession without progressive, penetrative actions can be low-value control that doesn't translate into chances.

For how clubs combine this data with tracking, video and injury-risk information across recruitment, coaching and the business side, see how sports teams use analytics and how to choose a sports performance analytics tool. Browse every tool in this category.