Football Analytics
Glossary

The probability that a given shot results in a goal, calculated from historical shot data. Factors include shot location, angle, body part used, whether it was a header, assist type, and game state. A shot from six yards out with no defender has a high xG (≈0.70); a 25-yard long-range attempt has a low xG (≈0.03).

Example: A striker scoring 12 goals on 8.4 xG is significantly overperforming. A striker scoring 7 on 10.1 xG may be underperforming, or facing poor finishing sequences that should regress to the mean.

The total xG of chances created by a player's passes, regardless of whether the recipient converted. xA separates the quality of a passer's delivery from a finisher's ability, making it more repeatable than actual assists as a measure of creative contribution.

Example: A player with 4 assists but 8.2 xA is generating far more high-quality chances than their assist count suggests, a strong indicator of sustained creative output.

The difference between a player's actual goal total and their xG total. Persistent large overperformance over multiple seasons suggests elite finishing technique; persistent underperformance may indicate poor decision-making in the final third or a regression risk.

Example: A forward consistently scoring 3 to 5 goals above xG across three seasons has a demonstrable finishing edge that standard shot quality models don't fully capture.

xG with penalties removed. Since penalties carry a fixed xG of approximately 0.76 regardless of the taker, including them distorts the comparison of strikers who earn different numbers of spot kicks. npxG gives a cleaner picture of open-play and set-piece quality.

The total number of offensive actions (passes, carries, dribbles, drawing fouls) that directly lead to a shot attempt. A broader measure of creative involvement than xA alone.

Ball carries that move the ball at least 10 metres closer to the opponent's goal, or into the opponent's penalty area. A high progressive carry rate per 90 indicates a player who can advance possession through dribbling, critical for wide players and advanced midfielders in counter-attacking systems.

Example: A winger with 6.4 progressive carries per 90 in the Bundesliga top quartile is a reliable ball-carrier even when passing options are limited.

Passes that advance the ball at least 10 metres towards the opponent's goal, or complete a pass from the defensive third into the middle third, or from outside the penalty area to inside it. A key indicator of a midfielder or centre-back's ability to play through lines.

A framework that assigns a numerical value to every on-ball action based on how much it changes the team's probability of scoring. Also known as On-Ball Value (OBV) or Expected Possession Value (EPV). Every pass, carry, tackle, and dribble is scored, enabling total contribution evaluation beyond goals and assists.

Example: A defensive midfielder with +18 OBV per season is contributing more to overall team success than a raw goals-and-assists reading would suggest.

Passes played into space behind the defensive line for a teammate to run onto. Through ball accuracy (successful through balls / attempted through balls) is a precise measure of a playmaker's vision and timing.

Total metres carried with the ball per 90 minutes. Distinguishes between high-volume dribblers operating in tight spaces and players who cover large distances with the ball in open space.

The average number of passes the opposition completes before a team makes a defensive action (tackle, interception, or foul) in the opponent's half. A lower PPDA indicates a more aggressive, higher-intensity press. Developed by Colin Trainor at StatsBomb.

Example: A team with PPDA of 6.2 is pressing very aggressively (roughly 6 passes before disruption); a team at 12.4 is pressing passively or sitting in a low block.

A composite measure of how actively a player participates in the team's press. Typically combines pressing actions per 90, PPDA contribution, and the percentage of pressing attempts that successfully force a turnover.

Specific situations or visual cues that prompt a player (or team) to initiate a press. Common triggers include a poor first touch by an opponent, a backwards pass, or an opponent receiving in a specific zone. Identifying a player's pressing trigger profile is a computer vision task.

How high a defensive unit positions their defensive line relative to their own goal. A high defensive line compresses space in midfield and supports the press; a low line protects against balls in behind but cedes midfield territory.

The percentage of contested aerial challenges won by a player. Important for centre-backs, target strikers, and players operating in direct, long-ball systems.

The percentage of attempted tackles that successfully win the ball. A high rate suggests precise defensive timing; a low rate on high tackle volume suggests recklessness.

A composite metric quantifying how well a player's statistical profile and playing style align with the behavioural fingerprint of a specific team system. Calculated by comparing player metrics, PPDA contribution, progressive carry rate, xG profile, pressing triggers, against the target system's profile. Outputs a ranked fit score with a metric-by-metric breakdown.

Example: A ball-carrying winger with high progressive carry rate and aggressive pressing stats scores highly for tactical fit in a gegenpressing system but poorly for a possession-focused 4-3-3 requiring patient rotational movement.

An analytical method using Bayesian distance metrics to identify the statistically closest players to a target profile across leagues, ages, and seasons. Enables the question "who plays like X but is younger or available?" to be answered computationally rather than through manual observation.

The use of statistical and machine learning models to score, rank, and filter potential transfer targets against defined criteria, system fit, developmental potential, cost-relative-to-quality, and positional need. Replaces subjective shortlisting with a documented, auditable process.

Expressing metrics per 90 minutes of playing time rather than per season totals. Allows fair comparison between players with different appearances, substitute roles, and match durations. Standard practice in professional football analytics.

Ranking a player's metric within a defined peer group (same position, same league, minimum minutes threshold) from 0 to 100. A percentile rank of 90 means the player outperforms 90% of their positional peers on that metric within the selected league and season context.

A regularised linear regression technique that prevents overfitting by penalising large model coefficients. Used in football analytics for development forecasting, predicting a player's future performance metrics from their current trajectory while avoiding overfit to noisy seasonal data.

A probabilistic clustering approach that groups players by statistical similarity using Bayesian distance metrics across multi-dimensional metric spaces. More robust than Euclidean distance methods for high-dimensional player data because it accounts for metric correlations and distributional differences across leagues.

A range within which the true value is expected to fall with a specified probability. In development forecasting, a 95% confidence interval on a predicted xG of 0.28 per 90 might be ±0.06, reflecting the genuine uncertainty in long-term player projection.

The use of computer vision and machine learning to categorise a player's playing style archetype from video footage, for example, "deep-lying playmaker", "inverted winger", "pressing centre-forward". Particularly valuable for youth players with insufficient statistical coverage.

A player's distinctive movement pattern tracked over a match: their runs without the ball, pressing routes, recovery shape, and off-ball positioning habits. Extracted via computer vision from tracking data or video analysis.