Match Analysis Fundamentals: Event Data, Tracking Data, and Tactical Metrics

Learning Objectives

After reading this article, you will be able to:

• Explain the definitions, collection methods, and characteristic differences between event data and tracking data.
• Understand the operating principles and validity-reliability limitations of major tracking systems (GPS/GNSS, optical tracking, LPS).
• Describe the definition, types, and design process of Key Performance Indicators (KPIs).
• Understand how contextual variables (match location, opposition quality, possession) influence match data interpretation.
• Explain the rationale and methods for integrating physical and tactical data in match analysis.

Three Pillars of Match Analysis Data

Football match analysis relies on two primary data types and the metrics derived from their combination. Event data records discrete actions that occur during a match — passes, shots, tackles, interceptions, fouls, and other identifiable behaviours. These data are collected by trained human coders, often employed by commercial providers such as Opta or StatsPerform, who tag each action with a timestamp, location, and outcome. Tracking data, by contrast, captures the continuous spatiotemporal coordinates of every player on the pitch, typically at rates of 10–25 samples per second. Tracking systems record where players are and how they move, generating raw position data from which velocity, acceleration, and distance metrics are derived (Clubb & Murray, 2022).

The collection processes behind these two data types are fundamentally different. Event data relies on human annotation guided by strict operational definitions: a pass is coded when possession is deliberately transferred between players, a shot when the ball is directed toward goal. This process inherits the strengths and limitations of notational analysis, a methodology with roots stretching back decades. Tracking data, meanwhile, is produced by technological systems — satellite-based GPS receivers, stadium-mounted cameras, or radio-frequency sensors — that automatically capture player positions without human judgment. A third category, tactical metrics, emerges when event and tracking data are combined or further processed. Expected Goals (xG), for example, models the probability of a shot resulting in a goal based on location, angle, and situational context. These derived metrics require both underlying data types and statistical modelling to produce.

The distinction matters because each data type captures different aspects of the match. Event data answers the question “what happened?” — how many passes were completed, where shots were taken, which players created chances. Tracking data answers “where and how much?” — total distance covered, time spent in different speed zones, acceleration and deceleration patterns. Neither type alone provides a complete picture. A player may cover enormous distances but contribute little tactically, or a team may execute few passes but create high-quality chances. A worldwide survey of over 200 football practitioners confirmed that locomotor variables — speed-zone-based distances — were the most commonly collected metrics across performance, data, and medical departments, used by 69–72% of respondents (Dello Iacono et al., 2025). Yet these movement variables tell us nothing about what the movement was for.

The central insight for practitioners entering this field is that event data, tracking data, and tactical metrics are complementary, not competing. Relying on any single data type produces an incomplete and potentially misleading account of match performance. The sections that follow examine the systems that generate these data, the metrics derived from them, and the principles required to interpret them meaningfully.

Tracking Technologies: GPS, Optical Tracking, and LPS

Three main technologies are used to generate tracking data in football. Global Navigation Satellite Systems (GNSS) — commonly referred to as GPS — use signals from orbiting satellites to calculate a receiver’s position on Earth. Players wear a small device between the shoulder blades, and the receiver triangulates its position using signals from at least four satellites. Modern football GPS units sample at 10 Hz or higher and include integrated inertial measurement units (IMUs) containing accelerometers, gyroscopes, and magnetometers. Optical Tracking Systems (OTS) use multiple cameras mounted around a stadium to record player positions. Computer vision algorithms process the video feed to extract x-y coordinates for every player, referee, and the ball, typically at 25 Hz. Local Positioning Systems (LPS) use radio-frequency signals — either Radio Frequency Identification (RFID) or Ultra-Wideband (UWB) — transmitted between player-worn tags and fixed antennae around the venue (Clubb & Murray, 2022).

A widely adopted classification organises the metrics these systems produce into three levels (Buchheit & Simpson, 2017). Level 1 metrics include distances covered in various speed zones — available from all tracking technologies. Level 2 metrics involve velocity changes, accelerations, decelerations, and direction changes — available from most systems but with increasing measurement uncertainty. Level 3 metrics are derived from inertial sensors embedded in wearable devices — available only from microtechnology units. GPS devices derive speed using two methods: positional differentiation (comparing successive coordinates) and the Doppler-shift method (measuring frequency changes in satellite signals), with the latter offering superior precision (Clubb & Murray, 2022).

Each system operates under different physical principles, and these differences affect measurement quality. A validation study comparing TRACAB Gen5 (an optical tracking system) against a VICON motion capture reference reported a position error of just 0.08 m and a total distance deviation of 0.27% (Linke et al., 2020). When a 10 Hz GPS device and a 25 Hz optical system were used simultaneously during official matches, both systems showed strong overall agreement for aggregate metrics. However, agreement weakened progressively as speed increased: the coefficient of variation between systems rose from less than 5% for walking and jogging to 14.9% for sprinting (Pons et al., 2019). This pattern — declining precision at higher velocities and during rapid changes of direction — is consistent across all tracking technologies (Murray & Clubb, 2022).

Feature	GPS/GNSS	Optical Tracking (OTS)	LPS (RFID/UWB)
Signal source	Satellites	Stadium cameras	Fixed antennae
Typical sampling rate	10–18 Hz	25 Hz	20–100+ Hz
Portability	High (wearable)	Low (fixed install)	Low (fixed install)
Ball tracking	No	Yes	Only with embedded sensor
Indoor use	No	Yes	Yes
Key limitation	Signal varies by environment	High cost; limited to venue	Infrastructure cost; signal interference

A critical practical consequence is that inter-unit variability within GPS devices of the same brand can reach up to 50%, meaning data from different units assigned to different players cannot be directly compared without accounting for device error (Buchheit & Simpson, 2017). Aggregate metrics such as total distance or high-speed running distance from one system cannot be directly pooled with data from another. Calibration equations have been developed to convert between systems, but these are specific to the manufacturer, model, and firmware version in use at the time of calibration (Murray & Clubb, 2022).

Validation is not a one-time event. Firmware updates, software changes, and environmental conditions (weather, satellite availability, stadium structure) all influence measurement accuracy. The best practice recommendation is to assign the same device to the same player across all sessions and treat data from different tracking systems as fundamentally non-interchangeable. Practitioners should also recognise a paradox identified in the literature: the variables most valued by practitioners — high-speed running distance, accelerations, decelerations, and metabolic power — are precisely the ones with the lowest measurement validity and reliability (Buchheit & Simpson, 2017).

Event Data: Passes, Shots, and the World of xG

Notational analysis is a systematic method for recording specific actions during competition. Early notational analysis was conducted manually with pen and paper. Today, commercial providers employ teams of trained coders who annotate every identifiable action in real time or from broadcast footage. The result is a dataset containing thousands of discrete events per match, each tagged with a timestamp, pitch coordinates, the players involved, and the action outcome. This is the origin of modern event data (Cardinale, 2022).

The value of event data lies in its ability to construct detailed technical performance profiles. A full-season analysis of 380 La Liga matches involving 409 outfield players examined 21 event-based variables across five positions (Liu et al., 2015). Players from stronger teams demonstrated higher involvement in attacking organisation — more ball touches, passes, through balls, and successful dribbles — while players from weaker teams performed more clearances and interceptions. Positional differences were equally pronounced: central attacking players led in shots and runs behind the defence, while central defenders led in clearances and covering actions. Critically, team quality and opposition quality had a greater influence on performance variability than match location or match outcome.

Beyond frequency counts, modern event data feeds increasingly sophisticated derived metrics. Expected Goals (xG) estimates the probability that a given shot will result in a goal, based on variables such as shot location, angle, body part, and defensive pressure. Expected Goals On-target (xGOT) refines this further by incorporating shot placement information. An analysis of 341 goals across the 2018 and 2022 FIFA World Cups used these metrics to evaluate scoring efficiency, finding that open-play scoring increased substantially between tournaments while the gap between expected and actual goals narrowed (Degrenne & Carling, 2024). These derived metrics transform raw event counts into probabilistic assessments, offering a more nuanced evaluation of attacking output than simple shot totals.

Event data is powerful for profiling and benchmarking, but it carries a critical limitation: high variability in key offensive metrics. Shots, shots on target, and assists exhibited the highest match-to-match variability across all contextual conditions examined (Liu et al., 2015). Evaluating a player’s attacking contribution based on a small number of matches is therefore inherently unreliable. Practitioners using these variables for player assessment must collect data across a sufficient number of matches before drawing conclusions.

A deeper methodological concern extends beyond variability. Frequency-based event counts do not capture the sequential context of a match — the order in which events occur, the interactions between players, or the tactical dynamics that produce a given outcome. A team may record many shots precisely because they are trailing and pushing forward recklessly, not because their attacking play is effective. This creates a risk of spurious correlation, where a statistical association between two variables reflects a shared cause rather than a direct relationship. A related problem is endogeneity: the outcome variable can influence the predictor. For example, fast breaks may appear to cause goals, but a team’s lead may itself provoke the opponent’s forward pressing that creates fast-break opportunities (Morgulev & Lebed, 2024). These issues mean that event data should always be interpreted within the broader tactical and situational context of the match.

Designing Tactical Metrics and KPIs

A Key Performance Indicator (KPI) is a quantitative measure used to evaluate whether a team or player is achieving defined performance goals. Three requirements distinguish a KPI from a general statistic: it must be measurable, objective, and actionable. KPIs cannot be designed in isolation. They must follow from clearly defined Performance Objectives (POs) — the specific outcomes a team or player is working toward, such as winning the league, qualifying for a tournament, or improving defensive structure (Cardinale, 2022).

The process of KPI development follows a structured workflow: identify the requirements for success, define a performance model that maps the determinants of that success, select KPIs that measure those determinants, assess the athlete or team, plan interventions, and review progress. KPIs in football span four domains: technical (pass completion rate, shot accuracy), tactical (pressing success rate, territorial control), physiological (high-speed running distance, sprint counts), and psychological (confidence, attentional control). A deterministic model can structure the relationship between performance outcomes and their contributing factors, although such models have been criticised for being overly rigid when applied to the nonlinear dynamics of team sports. Benchmarking individual KPI values typically relies on standardised scoring methods such as z-scores, which express a player’s performance as the number of standard deviations from the group mean. This allows practitioners to compare players across different metrics on a common scale and to identify specific strengths and weaknesses relative to peers (Cardinale, 2022).

A fundamental challenge with tactical KPIs is the assumption that experts agree on what they observe. A series of studies tested the level of agreement among football coaches analysing identical video footage (Furley et al., 2024). The results were striking: agreement on formation identification, spatial occupation, build-up play patterns, group tactical actions, and key player identification was consistently very low across three independent studies involving 15–24 coaches. This finding directly challenges the widely held belief that video analysis provides objective information. If experienced, qualified coaches watching the same footage arrive at substantially different tactical conclusions, the subjective component of tactical analysis must be acknowledged rather than assumed away.

At a more fundamental level, the dominance of KPI-centric research in football performance analysis has been questioned. Descriptive and correlational studies that count event frequencies and relate them to match outcomes face three structural problems: difficulty measuring truly meaningful predictor variables, the endogeneity problem described in the previous section, and a general absence of theoretical frameworks to guide hypothesis generation (Morgulev & Lebed, 2024). Alternative approaches — including network science, which models team interactions as dynamic networks, and dynamic systems analysis, which treats opponent-induced disruptions as perturbations to system equilibrium — offer paths beyond traditional KPI analysis. For practitioners, the practical implication is that KPIs remain useful as monitoring tools, but they should be designed in collaboration with coaching staff, anchored to specific Performance Objectives, and interpreted within a broader analytical framework rather than treated as standalone indicators of success.

Context Changes Data: Principles of Integrated Interpretation

Contextual variables are the situational factors that shape the physical, technical, and tactical outputs observed in match data. The most commonly studied contextual variables in football include match location (home or away), opposition quality (league ranking or competitive level), ball possession percentage, match status (leading, drawing, or trailing), and ball-in-play time. Ignoring these variables when interpreting performance data leads to systematic errors.

The influence of context on running data is dramatic. A five-season analysis of 1,675 English Premier League matches found that total running distance bore virtually no relationship with match success (Allen et al., 2026). Teams that ran more did not win more. However, when running data was separated into time in possession (TIP) and time out of possession (TOOP), a clear pattern emerged. The ratio of out-of-possession running intensity to in-possession running intensity showed a large positive association with points earned. Possession itself showed a very large positive relationship with match success. The top teams exhibited a characteristic physical signature: lower output during possession — reflecting positional control and structured play — combined with higher intensity out of possession, reflecting organised pressing and rapid ball recovery. This pattern was also sensitive to opposition quality: when facing a top-six opponent, both top and lower-ranked teams adjusted their physical profiles, with possession dropping and ball-in-play time increasing.

The integration of physical and tactical data takes contextualisation further. An analysis of 50 Premier League matches involving 244 players coded each instance of high-intensity running (above 19.8 km/h) according to 11 specific tactical actions — pressing, covering, recovery runs, overlapping, breaking into the box, and others (Ju et al., 2023). The results revealed that using broad positional categories systematically distorted the picture. Full-backs covered 34% less high-intensity distance than the wider “wide defensive player” category suggested, while wing-backs covered 15% more. Similarly, central defensive midfielders ran 30% less at high intensity than the generic “central midfielder” average, and pressing activity per match was just 2 instances for defensive midfielders compared with 9 for attacking midfielders — yet the general category averaged 5. These differences directly affect training prescription, physical benchmarks, and player recruitment decisions.

The validity and reliability of this integrated physical-tactical approach have been formally established. Across 30 expert evaluators (UEFA-qualified coaches and performance analysts), overall classification accuracy reached 91.8%, with strong inter-observer agreement and almost perfect intra-observer agreement (Ju et al., 2022). This provides confidence that the approach can be applied systematically, though it remains labour-intensive.

The concept of purposeful distances captures the essential principle. Raw tracking data produces what has been termed “one-dimensional and blind distances” — numbers stripped of tactical meaning (Murray & Clubb, 2022). A player who covers 12 km in a match may have spent that effort pressing effectively, recovering defensively, or simply jogging out of position. Without the tactical overlay, the number is ambiguous at best and misleading at worst. Integrating physical data with tactical context transforms blind distances into purposeful ones, connecting movement to its tactical intent.

The broader lesson is that technology in match analysis is valuable only when it provides information that cannot be obtained through simpler means. The ultimate value comes not from the systems or the metrics themselves, but from the practitioners who interpret, contextualise, and act on the data (Buchheit & Simpson, 2017). As tracking systems become more accessible and data volumes grow, the ability to ask the right questions of the data — and to resist the temptation to treat numbers as self-explanatory — becomes the most important skill in the analyst’s toolkit.

Key Takeaways

• Event data records discrete match actions (passes, shots, tackles) while tracking data continuously records players’ spatiotemporal coordinates — the two are complementary and incomplete in isolation.
• GPS, optical tracking, and LPS each have unique strengths and weaknesses, and all systems lose precision during rapid velocity changes (accelerations, decelerations, direction changes) — data from different systems must not be directly compared.
• KPIs should be established only after Performance Objectives are defined and must meet three requirements — measurability, objectivity, and actionability — KPIs designed without coaching staff collaboration risk going unused.
• Contextual variables such as match location, opposition quality, and possession significantly influence the incidence of physical and technical performance data, so interpreting numbers without context leads to errors.
• Integrating tracking data with tactical actions transforms ‘blind distances’ into ‘purposeful distances’, providing practical implications for position-specific training design and player recruitment.

References

1. Allen, T., Taberner, M., Zhilkin, M., Harper, D., & Alexander, J. (2026). Possession in motion a five-season analysis of running in-possession and out-of-possession with match outcomes in the English Premier League. Biology of Sport. https://doi.org/10.5114/biolsport.2026.159531
2. Buchheit, M. & Simpson, B. M. (2017). Player-Tracking Technology: Half-Full or Half-Empty Glass?. International Journal of Sports Physiology and Performance, 12(s2), S2-35-S2-41. https://doi.org/10.1123/ijspp.2016-0499
3. Cardinale, M. (2022). Key Performance Indicators. In D. N. French & L. Torres Ronda (Eds.), NSCA’s Essentials of Sport Science. Human Kinetics.
4. Clubb, J., & Murray, A. M. (2022). Characteristics of tracking systems and load monitoring. In D. N. French & L. Torres Ronda (Eds.), NSCA’s Essentials of Sport Science. Human Kinetics.
5. Degrenne, O. & Carling, C. (2024). Comparison of goalscoring patterns between the 2018 and 2022 FIFA World Cups. Frontiers in Sports and Active Living, 6, 1394621. https://doi.org/10.3389/fspor.2024.1394621
6. Dello Iacono, A., Datson, N., Clubb, J., Lacome, M., Sullivan, A., & Shushan, T. (2025). Data analytics practices and reporting strategies in senior football: insights into athlete health and performance from over 200 practitioners worldwide. Science and Medicine in Football, 10(1), 80-95. https://doi.org/10.1080/24733938.2025.2476478
7. Furley, P., Mehta, S., Raabe, D., & Memmert, D. (2024). Objectivity of match analysis in football: Testing the level of agreement between coaches’ interpretations of video data. International Journal of Sports Science & Coaching, 20(1), 45-55. https://doi.org/10.1177/17479541241278603
8. Ju, W., J Lewis, C., Evans, M., Laws, A., & S Bradley, P. (2022). The validity and reliability of an integrated approach for quantifying match physical-tactical performance. Biology of Sport, 39(2), 253-261. https://doi.org/10.5114/biolsport.2022.104919
9. Ju, W., Doran, D., Hawkins, R., Evans, M., Laws, A., & Bradley, P. (2023). Contextualised high-intensity running profiles of elite football players with reference to general and specialised tactical roles. Biology of Sport, 40(1), 291-301. https://doi.org/10.5114/biolsport.2023.116003
10. Linke, D., Link, D., & Lames, M. (2020). Football-specific validity of TRACAB’s optical video tracking systems. PLOS ONE, 15(3), e0230179. https://doi.org/10.1371/journal.pone.0230179
11. Liu, H., Gómez, M., Gonçalves, B., & Sampaio, J. (2015). Technical performance and match-to-match variation in elite football teams. Journal of Sports Sciences, 34(6), 509-518. https://doi.org/10.1080/02640414.2015.1117121
12. Morgulev, E. & Lebed, F. (2024). Beyond key performance indicators. German Journal of Exercise and Sport Research, 54(3), 335-340. https://doi.org/10.1007/s12662-024-00944-8
13. Murray, A. M., & Clubb, J. (2022). Analysis of tracking systems and load monitoring. In D. N. French & L. Torres Ronda (Eds.), NSCA’s Essentials of Sport Science. Human Kinetics.
14. Pons, E., García-Calvo, T., Resta, R., Blanco, H., López Del Campo, R., Díaz García, J., & Pulido, J. J. (2019). A comparison of a GPS device and a multi-camera video technology during official soccer matches: Agreement between systems. PLOS ONE, 14(8), e0220729. https://doi.org/10.1371/journal.pone.0220729