by NSCA’s Essentials of Sport Science
Kinetic Select March 2022
The following is an exclusive excerpt from the book NSCA’s Essentials of Sport Science, published by Human Kinetics. All text and images provided by Human Kinetics.
The latest models of most GPS devices appear to provide an adequate measure of distance and speed (10, 74). Therefore, sport scientists can have relative confidence in associated metrics (i.e., total distance covered and the breakdown of the distance within various speed thresholds). However, a familiar pattern in GPS validation research, as with positional systems, suggests that the precision of GPS to measure speed and distance decreases with an increased change in the rate of velocity, regardless of the sampling rate (1, 74). Validation research has associated a higher sampling frequency with an increased accuracy of GPS devices (74). While this may be true for 10-Hz versus 1- to 5-Hz devices, research comparing 10-Hz and 18-Hz devices has shown mixed results. One study showed no meaningful differences between the two sampling frequencies, whereas another demonstrated improved validity and reliability in the 18-Hz compared to the 10-Hz device (10, 37). It is worth noting that differences exist in the satellite systems that the respective technologies have access to, with the increased sampling frequency of 18 Hz used in devices that have access only to the United States GPS (10). This is opposed to the 10-Hz multi-GNSS device within that study, which is capable of tracking multiple satellite systems including both GPS and Global Navigation Satellite System (GNSS) (10). While more research is necessary regarding the relative advantages of devices operating at higher sampling frequencies (>10 Hz), practitioners should have confidence in devices that operate at or above this threshold.
Rather than high absolute velocity, it is typically the actions that involve a high rate of change in velocity (i.e., acceleration, deceleration, and change of direction) that provide moments crucial to a game and its outcome. The measurement of these actions may, however, be associated with the greatest amount of error. Measuring instantaneous acceleration is a mathematical ideal limited by the sampling frequency of the device. Mathematically, as the denominator in the equation approaches zero, the limit of the rate is called a derivative. In this case acceleration (a) is a derivative of velocity (v):
Acceleration is the derivative of velocity with time (t), but velocity is itself the derivative of displacement (s) with time (38). Therefore, acceleration is the first derivative of velocity with time and the second derivative of displacement with time:
Calculating from the second derivative can compound calculated errors but is typically the way acceleration is calculated in GPS units that measure geographical position. Noise within the raw signal may be amplified (56, 81). For these reasons, positional data are often filtered to remove high-frequency noise (42, 56, 61).
Knowing the acceleration over time, one can work backward to calculate the integral of change in velocity via the first equation of motion (21, 38):
Another consideration for sport scientists using GPS data in their monitoring practices is the signal quality. This can be affected by local weather, the location (e.g., within a densely populated area versus a remote location), and environmental obstructions such as trees close to training locations or particular stadium designs (e.g., partially covered) (46). Signal quality can be evaluated by quantifying the satellite number used to obtain the signal or the horizontal dilution of precision (HDoP) or both. The HDoP represents the precision of the GPS horizontal positional signal, determined by the satellites’ geographical organization (46). Values range from 1 to 50, representing a ratio of the change in output location and the change in measured data, with 1 being ideal and under 2 excellent, whereas values over 20 are regarded as poor (82). Low values represent satellites far apart in the sky, with a wide angular separation giving the best positional information. While considering signal quality is a best practice in cleaning the GPS data, not all manufacturers make it available in their software. This can make it difficult to identify and potentially compensate for sections of poor signal quality within the collected data.
Drawing definitive conclusions on validity for tracking systems is problematic, given the variation in manufacturers, models, sampling frequencies, sport demands, variables assessed, thresholds used, and software and hardware versions (14). Specifically related to GPS technology, there are also many factors that can influence the validity, including the method used to calculate distance and velocity, the signal processing algorithms, the GNSS available, and the chipset technology used. Since these factors differ in GPS devices depending on the manufacturer, the GPS device model (within manufacturer), and the firmware version used (within-device model), the validation of these devices remains an ongoing process.
Sport scientists should consider if a suitable criterion measure was used to assess the GPS device when elucidating the results of validation research. When assessing speed, appropriate criterion measures include a laser, radar, or 3D-motion capture system. Sport scientists should be attentive if timing gates are used to assess speed, since this will not provide a measure of instantaneous speed, but rather an average. When assessing distance, appropriate criterion measures include a theodolite, tape measure, trundle wheel, and 3D-motion capture systems. The movement scenario used in the methodology (i.e., predefined course versus complex and free movement) should also be considered.
The reliability of GPS devices is more complex to assess than validity. Within-device reliability would require nearly identical repetition of specific movement patterns. However, it is unlikely that human participants will produce with 100% accuracy the exact same movement patterns on multiple occasions throughout a sport-specific circuit (46). Therefore, most reliability research has investigated inter-unit (between units of the same model within manufacturer) and inter-device (between manufacturers) reliability. The 10-Hz GPS models from three different manufacturers have been compared for inter-unit and inter-device reliability when data was processed using manufacturer software (70). Inter-unit reliability was found to be poor for deceleration measures (coefficient of variation [CV]: 2.5%-72.8%), better with acceleration measures (CV: 1.4%-19.7%), and best for distance and speed measures (CV: 0.2%-5.5%). However, significant differences were observed between manufacturers for all measures (CV: −2.0% to 1.9%). While it appears that there may be some flexibility in device interchangeability using the same model, depending on the measures of interest to the sport scientist, the best practice of using the same device on a given player should still be applied where possible (39). It is apparent that data derived across manufacturers are not interchangeable, and caution is warranted when one is comparing such data.
In the applied setting of sport training and competition, it is commonplace for sport scientists to employ real-time monitoringof GPS data in order to provide the coaches or players (or both) with feedback relating to the volume and intensity of a session (46). Such feedback may be based on post-event data, downloaded directly from the device, whereas real-time data is derived via a specific receiver (6). Therefore, understanding the agreement between these data sets is essential. An early study in this regard published in 2010 (GPS frequency not specified, but thought to be 1 or 5 Hz) questioned the application of real-time data for decision making, given that only total distance displayed a signal greater than the noise (4). However, technology has developed since that study, and therefore two more investigations have reexamined the validity. This research has disagreed on the validity of real-time, accelerometry-derived data (6, 79). As with post-event velocity data, there is consensus that error increases with higher speed (see figure 10.1), but these data may still be used if the error is taken into account appropriately (6, 79). Given the paucity and inconsistencies of research regarding real-time monitoring, along with the variation in software and hardware available, conducting analysis within the applied environment that replicates the methodologies used in the literature represents best practice for practitioners using real-time data to assist with feedback and decision making.
NSCA’s Essentials of Sport Science provides the most contemporary and comprehensive overview of the field of sport science and the role of the sport scientist. It is a primary preparation resource for the Certified Performance and Sport Scientist (CPSS) certification exam. The book is available in bookstores everywhere, as well as online at the NSCA Store.