Tracking Technology in Football: Comparing GPS, Optical, LPS, and IMU Systems

Prerequisites: This article assumes familiarity with external and internal training load concepts and a basic understanding of GPS and accelerometer tracking systems. If any of these topics are new to you, start with:

• External and Internal Load Concepts and Monitoring
• Understanding GPS, Accelerometers, and Tracking Systems

Learning Objectives

• Distinguish the operating principles and data output methods of GPS/GNSS, Optical Tracking Systems (OTS), RFID/LPS, and Inertial Measurement Units (IMU).
• Understand the validity and reliability characteristics and measurement limitations of each tracking system.
• Explain between-systems agreement issues and the need for calibration equations.
• Understand the metric classification framework (Level 1/2/3) for outputs derived from tracking systems.
• Apply criteria for selecting tracking technology based on environment and purpose.

Evolution of Player Tracking: From Manual Coding to Automated Systems

The first objective measurement of player movement in football dates back to 1976, when Tom Reilly used hand-notation coding systems to record time-motion data during matches (Clubb & Murray, 2022). A single analyst would watch one player for the entire game, manually logging each movement into speed categories. The method was time-consuming, restricted to one player at a time, and produced only coarse movement summaries.

Technological progress has since enabled simultaneous, automated tracking of every player on the pitch. Modern tracking methods fall into two broad categories: positional systems and wearable microsensors (Clubb & Murray, 2022). Positional systems — including Optical Tracking Systems (OTS), Local Positioning Systems (LPS) based on Radio Frequency Identification (RFID), and the Global Positioning System (GPS) — determine a player’s location through external infrastructure such as cameras, anchor nodes, or satellites. Wearable microsensors — primarily Inertial Measurement Units (IMU) containing accelerometers, gyroscopes, and magnetometers — measure movement directly from the player’s body.

In practice, the two categories are increasingly combined. A single GPS unit worn between the shoulder blades now typically houses a GNSS receiver, a tri-axial accelerometer, a gyroscope, and a magnetometer (Clubb & Murray, 2022). Understanding what each sensor measures — and where its measurement breaks down — is the foundation for interpreting the data these systems produce.

GPS/GNSS: Principles and Limitations of Satellite-Based Tracking

GPS calculates a player’s position by measuring the time delay between signals transmitted from orbiting satellites and received by a wearable unit. A minimum of four visible satellites is required to solve for three-dimensional position and time. The term Global Navigation Satellite System (GNSS) is the broader designation that includes the United States GPS constellation (24 satellites) and Russia’s GLONASS (24 satellites); modern receivers typically access both networks simultaneously (Clubb & Murray, 2022).

Two methods derive speed and distance from satellite signals. Positional differentiation calculates displacement between consecutive position fixes over time. The Doppler-shift method measures the frequency change of the satellite signal caused by receiver movement. The Doppler-shift method provides higher speed precision, and team sport GPS devices typically use it for velocity output, while distance is derived from positional differentiation (Clubb & Murray, 2022).

Sampling rate affects measurement capability. Early devices operated at 1 Hz; the current recommendation is 10 Hz or higher (Murray & Clubb, 2022). Some devices interpolate data from a lower native sampling rate using accelerometer fusion, inflating the reported frequency without increasing genuine GPS resolution. Practitioners should verify the true GNSS chip frequency before purchase.

Signal quality depends on the number of visible satellites and environmental conditions. Horizontal Dilution of Precision (HDoP) quantifies the geometric quality of satellite positions relative to the receiver. A value of 1 is ideal; values above 20 indicate poor accuracy. Stadium roofs, tall stands, and surrounding vegetation all degrade signal reception (Murray & Clubb, 2022).

Inter-unit variability is a critical concern. Differences between individual GPS units of the same brand and model can reach up to 50% for certain variables (Buchheit & Simpson, 2017). Assigning the same unit to the same player for every session is considered best practice, and a larger threshold for meaningful change should be applied when comparing data across players (Murray & Clubb, 2022).

The measurement precision hierarchy across variables follows a consistent pattern. Total distance and average speed are measured with acceptable accuracy (CV typically below 5%). As the rate of velocity change increases — through accelerations, decelerations, and direction changes — error rises sharply. A comparison of three 10 Hz GPS manufacturers found deceleration CV values ranging from 2.5% to 72.8%, while distance and speed CV remained between 0.2% and 5.5% (Murray & Clubb, 2022). This pattern is not unique to GPS. It is shared across all positional tracking technologies.

Camera-Based Tracking: Structure and Precision of Optical Systems

Optical Tracking Systems (OTS) reconstruct player movement by inferring x-y pitch coordinates from multiple cameras mounted around a stadium. A minimum of two camera viewpoints must cover every area of the pitch. Machine learning and computer vision algorithms identify each player and track their positions at sampling rates of 25 Hz or higher (Clubb & Murray, 2022).

Semi-automatic multi-camera systems were introduced to football in the late 1990s and enabled the first simultaneous tracking of all players, referees, and the ball across a full match (Clubb & Murray, 2022). Ball tracking remains a defining advantage: OTS generates tactical outputs — possession time, passing networks, pressing intensity — that no wearable device can produce. This capability also makes OTS applicable in sports where wearable devices are prohibited during competition.

Validation against a VICON motion capture reference demonstrated that the TRACAB Gen5 system achieves a position RMSE of 0.08 m, velocity RMSE of 0.08 m/s, and acceleration RMSE of 0.21 m/s² (Linke et al., 2020). All key performance indicators showed trivial deviations from the reference. The earlier Gen4 system produced small-to-moderate deviations in low-speed and high-speed distance categories, with a tracking error rate of 1.08% compared to 0% for Gen5 (Linke et al., 2020).

OTS carries practical limitations. Systems require permanent installation and dedicated computer networks, making them expensive and immobile. Vertical movement — jumping, heading duels — is not captured. In congested scenes such as corner kicks or goal-mouth scrambles, tracking errors from player occlusion can increase, though this scenario was not included in the TRACAB validation protocol (Linke et al., 2020). Post-processing delays of up to 24–36 hours in some semi-automated systems restrict real-time application during training (Clubb & Murray, 2022).

Local Positioning Systems: RFID and Ultra-Wideband Technology

Local Positioning Systems (LPS) calculate a player’s location by measuring the time-of-arrival of radio signals between fixed anchor nodes installed around a venue and a microchip transmitter worn by the player. Because LPS does not depend on satellite signals, it performs equally well indoors and outdoors (Clubb & Murray, 2022).

Position estimation accuracy for RFID-based systems ranges from 11.9 to 23.4 cm, and average acceleration and deceleration measurements fall within 2% of 3D motion capture values (Murray & Clubb, 2022). Some implementations operate at 100 Hz or higher, significantly exceeding the 10–18 Hz typical of GPS devices. This higher sampling rate provides better resolution for capturing rapid velocity changes.

RFID systems are susceptible to electronic interference and signal instability. Battery drain during extended sessions can produce outlier data, and outlier rejection rates have been reported higher than for GPS in some studies (Murray & Clubb, 2022). Maximum acceleration and deceleration measurements show lower reliability than average values, and accuracy degrades as directional complexity increases.

Ultra-Wideband (UWB) technology addresses several RFID limitations. UWB uses radio frequency signals with bandwidths exceeding 500 MHz, offering lower power consumption, higher precision, stronger obstacle penetration, and better resistance to electromagnetic interference (Clubb & Murray, 2022). UWB represents the current direction of LPS development for both indoor and hybrid tracking environments.

The main practical barrier for LPS is infrastructure. Anchor nodes require installation and calibration at each venue, and the system cannot travel to away fixtures. Ball tracking is only available if the ball carries an embedded sensor. These factors limit adoption to organisations with dedicated, permanent facilities and sufficient budgets.

Wearable Microsensors: Accelerometers, Gyroscopes, and Magnetometers

Inertial Measurement Units (IMU) combine three sensor types within a single Microelectromechanical Systems (MEMS) chip. Tri-axial accelerometers measure linear acceleration across x, y, and z axes at sampling rates exceeding 1,000 Hz. Gyroscopes detect angular velocity — the rate of rotation — around three axes. Magnetometers use the Earth’s magnetic field to determine heading and orientation, functioning as a digital compass (Clubb & Murray, 2022).

These sensors are complementary. Magnetometers lose accuracy during fast movements, while gyroscopes accumulate drift error over time. Combining both produces low-drift, rapid orientation estimation. Gyroscope data has also been applied to collision detection in contact sports, informing automated “tackle algorithms” for event identification (Clubb & Murray, 2022).

Accelerometer-derived metrics include Player Load (and equivalents such as Body Load and Accumulation Load), a scaled vector magnitude of tri-axial acceleration changes expressed in arbitrary units. It reflects the instantaneous rate of change in acceleration — a proxy for mechanical stress — rather than a direct physiological cost (Clubb & Murray, 2022). Beyond composite load, accelerometers provide insights into stride characteristics, bilateral asymmetry, ground contact time, and estimated ground reaction forces. These variables hold potential for longitudinal neuromuscular fatigue monitoring and injury risk tracking (Buchheit & Simpson, 2017).

A key limitation is sensor placement. Standard wearable devices sit between the shoulder blades, not at the body’s centre of mass. Player posture during acceleration, defensive crouching, or sport-specific body positions introduces systematic bias into accelerometer readings (Murray & Clubb, 2022). Within-device and between-device reliability are generally acceptable (CV approximately 1–2%), but the biological interpretation of accelerometer outputs depends on understanding the constraints imposed by the sensor’s fixed position on the upper back.

What Can Tracking Data Measure: The Three-Level Metric Classification

Not all tracking outputs carry equal measurement quality. A three-level classification framework organises the variables derived from tracking systems by signal source, measurement complexity, and the technology required to produce them (Buchheit & Simpson, 2017).

Level	Variables	Technology Required
Level 1	Distance covered by speed zone	All positional systems
Level 2	Velocity changes, acceleration, deceleration, change of direction	Most positional systems
Level 3	Inertial sensor events (impacts, Player Load, stride metrics)	Wearable microsensors only

Level 1 variables — total distance and distance by speed zone — are produced by any positional tracking system. They are the most reliably measured and the most frequently reported: total distance appeared in 47 of 82 studies in a systematic review of load monitoring in football (Miguel et al., 2021).

Level 2 variables include instantaneous velocity, acceleration, deceleration, and change-of-direction events. These require higher sampling rates and more sophisticated signal processing. Most GPS, OTS, and LPS systems generate Level 2 data, but measurement error is substantially higher than for Level 1. Acceleration is the first derivative of velocity (itself derived from position), meaning errors compound at each differentiation step (Murray & Clubb, 2022). The challenge is amplified by the use of arbitrary absolute thresholds. A fixed high-intensity acceleration threshold of 3 m/s² is common, yet players’ maximal acceleration capacity can exceed this value by 67% or more — meaning the same absolute number represents fundamentally different physiological demands for different athletes (Pimenta et al., 2026).

Level 3 variables are exclusive to wearable microsensors: Player Load, impact events, stride length and asymmetry, and estimated vertical stiffness. Because they originate from sensors on the player’s body rather than from external positional tracking, they capture movement characteristics independent of tactical context. This independence may make them more sensitive to fitness and fatigue changes than position-derived metrics, which are heavily influenced by tactical role, coaching instructions, and match state (Buchheit & Simpson, 2017).

A practical paradox emerges from this classification. The variables that practitioners consider most informative — high-speed running distance, acceleration and deceleration counts, metabolic power — sit at Level 2, where validity and reliability are lowest. Level 1 variables are measured most accurately but provide the least nuanced picture of training stimulus. Level 3 variables hold the greatest promise for individualised monitoring but still require further validation of their sensitivity to meaningful physiological changes (Buchheit & Simpson, 2017).

Different Systems, Same Player: Is Data Interchangeable?

Professional teams frequently use different tracking systems for training and competition. GPS devices are worn during training sessions, while league-mandated optical tracking captures match data. This dual-system reality is widespread and raises a practical question: can output from one system be compared with output from another (Riboli et al., 2023; Rice et al., 2023)?

A simultaneous comparison of a 10 Hz GPS device and a 25 Hz optical tracking system during 42 official matches found high overall agreement, with intraclass correlation coefficients exceeding 0.90 for all variables (Pons et al., 2019). Total distance, distance per minute, and low-speed zones showed coefficients of variation below 5%. Agreement deteriorated at higher speed thresholds: intense running (18–21 km/h) produced a CV of 10.4%, low-intensity sprinting (21–24 km/h) reached 13.5%, and high-intensity sprinting (>24 km/h) climbed to 14.9% (Pons et al., 2019). The optical system slightly overestimated distance variables, while the GPS device reported higher speed values.

A broader comparison of four systems during full football matches revealed larger discrepancies. When validated against 3D motion capture, RFID achieved the highest positional accuracy (distance RMSE 22–27 cm), but all technologies showed increased error at higher speeds and during rapid direction changes. Across key performance indicator categories, system-dependent variation exceeded practical interchangeability thresholds (Murray & Clubb, 2022).

Calibration equations offer a partial solution. Regression equations derived from simultaneous GPS-OTS measurement allow conversion of one system’s output into the other’s scale (Pons et al., 2019). These equations enable integrated week-to-week load monitoring when training and match data originate from different sources (Riboli et al., 2023). However, calibration equations are specific to the exact system pair, model, firmware version, and data processing pipeline. A software update from either manufacturer can invalidate the conversion.

The practical recommendation is unambiguous. Each organisation should conduct its own internal validation using its specific systems and conditions. Aggregate metrics from different systems should never be directly compared or combined without calibration (Murray & Clubb, 2022).

Which Tracking Technology Fits Our Team?

Technology selection depends on four primary factors: venue availability, budget, the depth of information required, and whether the environment is indoor or outdoor (Clubb & Murray, 2022).

Criterion	GPS/GNSS	OTS	LPS (RFID/UWB)	IMU (standalone)
Outdoor use	Yes	Yes	Yes	Yes
Indoor use	No	Yes	Yes	Yes
Ball tracking	No	Yes	Sensor required	No
Portability	High	Low	Low	High
Sampling rate	10–18 Hz	25+ Hz	100+ Hz	1,000+ Hz
Relative cost	Moderate	High	High	Low

GPS/GNSS offers the broadest accessibility. It works at any outdoor venue, travels with the team, and provides Level 1 and Level 2 data at moderate cost. Its limitations — satellite dependency, inter-unit variability, reduced accuracy at high speeds — are well-documented and manageable within a consistent internal protocol. OTS delivers the highest ecological validity for match analysis, including ball tracking and tactical outputs, but requires permanent stadium installation. LPS achieves the highest positional sampling rates and works indoors, making it suitable for enclosed-venue sports, but demands dedicated infrastructure. Standalone IMU devices are the most portable and affordable option, though they provide only Level 3 data without position-derived metrics.

Implementing a new system should follow a structured innovation process: awareness of a performance gap, evaluation of available solutions, a trial period under operational conditions, and adoption with ongoing validation (Torres Ronda, 2022). A common mistake is skipping directly to adoption — purchasing a system because competitors use it — without completing the evaluation and trial stages.

Athlete buy-in is a frequently overlooked success factor. The most advanced system is worthless if players do not wear devices consistently. Wearable units should be unobtrusive, comfortable, and ideally unnoticeable. Communicating to players what the data is used for — and how it benefits their preparation — is essential for sustained compliance (Torres Ronda, 2022). The guiding principle is not to acquire the most sophisticated technology, but to use the right technology at the right time in the right way, and to apply the knowledge it generates to practice (Torres Ronda, 2022).

Key Takeaways

• GPS/GNSS tracks players via satellite signals, OTS uses multi-camera arrays, LPS (RFID/UWB) uses anchor node radio signals, and IMU uses inertial sensors — each operates on fundamentally different principles and produces different output types.
• Across all tracking systems, measurement error increases at higher speeds; acceleration, deceleration, and change-of-direction variables carry the lowest measurement precision.
• Directly comparing aggregate metrics from different tracking systems is not recommended; calibration equations are necessary for between-system data integration, and each organisation should develop and validate these internally.
• The three-level metric classification (Level 1: distance by speed zone; Level 2: velocity and acceleration changes; Level 3: inertial sensor-derived events) maps which variables each technology can produce, with a trade-off between information value and measurement reliability.
• Technology selection should consider venue availability, budget, information needs, and indoor/outdoor environment, following a four-stage innovation process: awareness, evaluation, trial, and adoption.
• For GPS, a sampling rate of 10 Hz or above is recommended, and the Doppler-shift method provides more precise speed output than positional differentiation.
• Assigning the same unit to the same player for every session is best practice; data between different manufacturers, models, and firmware versions are not interchangeable.

References

1. 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
2. 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.
3. 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
4. Miguel, M., Oliveira, R., Loureiro, N., García-Rubio, J., & Ibáñez, S. J. (2021). Load Measures in Training/Match Monitoring in Soccer: A Systematic Review. International Journal of Environmental Research and Public Health, 18(5), 2721. https://doi.org/10.3390/ijerph18052721
5. 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.
6. Pimenta, R., Antunes, H., Silva, H., Ribeiro, J., & Nakamura, F. Y. (2026). The Need for GPS Data to be Normalized for Performance and Fatigue Monitoring in Soccer: Considerations for Accelerations and Decelerations. Strength & Conditioning Journal. https://doi.org/10.1519/ssc.0000000000000958
7. 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
8. Riboli, A., MacMillan, L., Calder, A., & Mason, L. (2023). Player monitoring and practical application. In A. Calder & A. Centofanti (Eds.), Peak performance for soccer: The elite coaching and training manual. Routledge.
9. Rice, J., Kovacevic, D., Calder, A., & Carter, J. (2023). Wearable technology. In A. Calder & A. Centofanti (Eds.), Peak performance for soccer: The elite coaching and training manual. Routledge.
10. Torres Ronda, L. (2022). Technological implementation. In D. N. French & L. Torres Ronda (Eds.), NSCA’s Essentials of Sport Science. Human Kinetics.