CC BY 4.0 (Creative Commons Attribution 4.0 International)Johnston, Richard (Principal Supervisor)Duthie, Grant (Co-Supervisor)Cole, Michael (Co-Supervisor)Crang, Zachary2025-11-122025-11-122025-11-1210.26199/acu/38151https://doi.org/10.26199/acu/38151https://hdl.handle.net/20.500.14802/38151The quantification of a player’s external training load (i.e., work performed) is important and is commonplace to numerous levels of team sport. These data are often used to establish match demands of a sport; and then incorporated into a feedback loop where they are used to: (1) prescribe training that is specific to those demands; (2) monitor training loads to ensure the desired outcomes are achieved; and (3) adjust loads for subsequent training periods, which can vary from within a session, to between seasons. All these steps are used to minimise injury risk, optimise adaptations and deliver successful performances. Historically this tracking was performed by charting player movements during match-play live or retrospectively. Semi- and fully-automated computer-vision tracking systems were then used but are labour intensive and therefore not suitable for the fast-paced nature of modern day sport. Hence the introduction of wearable microtechnology in the field of sport science has become the most common means of quantifying player movements. Recently, computer-vision and artificial intelligence (AI) have experienced rapid growth in the area of tracking player movements. Given the widespread use of these technologies, it is important to ensure their output is understood to facilitate effective use in practice. First and foremost, the validity and reliability of these devices must be understood. As such, the overall aim of this thesis was to establish the validity and reliability of common tracking technologies used in team sports. This thesis comprises a systematic review and three experimental studies. The systematic review appraised studies that investigated the validity and/or reliability of the wearable microtechnology devices used in team sports. After assessing the eligibility of the 384 retrieved studies, 72 were eligible and therefore included in the review. Global navigation satellite systems (GNSS) were examined in 47 studies, local positioning systems (LPS) in 12 studies and inertial measurement units (IMUs) in 25 studies. It was difficult to collectively synthesise the validity and reliability of wearable microtechnologies, given the methodological heterogeneity among the included studies. However, whilst validity and reliability varied across studies, in general there was a trend for improved validity and reliability as sampling frequency increased. Typically, the devices offered suitable accuracy for monitoring key metrics such as peak speed and distance covered. Significant gaps identified in the literature included: 1) the validity and reliability of the most recent and commonly used GNSS devices had not yet been examined in peer-reviewed research; 2) Intra-device reliability had been inadequately assessed; and 3) Validity and reliability had not yet been assessed over multiple days. Given the findings of the systematic review, Chapter 4 investigated the inter-device and inter-manufacturer reliability of Catapult (Vector S7) and Statsports (Apex Pro) devices over time. It was found that the devices possessed consistently suitable reliability for most metrics apart from threshold-based acceleration and deceleration metrics. There were also large differences between the outputs of both manufacturers. Chapter 5 assessed the validity of the same devices to measure instantaneous speed and acceleration during straight-line sprinting across multiple sessions. It was found that both devices possessed suitable validity for measuring speed and acceleration compared to a laser criterion. There were small differences in validity across sessions for Catapult units, but these were not practically meaningful. Chapter 6 assessed the ability of computer-vision and AI models to track players position and speed using broadcast video footage. The results showed that players can be accurately tracked but this is dependent upon the computer-vision and AI techniques implemented. Overall, modern day (≥10-Hz) GNSS devices provide suitable validity and reliability. Further, it appears that computer-vision tracking may be a suitable alternative provided players are detected by the software. However, future research should focus on this area given it is in its infancy. In particular, the methods used to estimate a player’s position when they are not detected by the software (e.g., outside of the camera’s field of view, occluded) currently lack suitable accuracy and should be further developed. It may be that more training data are required for the machine learning algorithms to develop sufficient accuracy to predict player position accurately in these situations. Collectively, this thesis consolidates the work pertaining to the validity and reliability of wearable player tracking devices. It provides practitioners with an understanding of the precision and accuracy of the metrics commonly assessed using wearable devices and highlights the viability of using computer-vision and AI to track players using broadcast footage.enplayer tracking technologiesteam sportsGlobal Navigation Satellite SystemsGNSSAIbroadcast footageperformance analysisathlete monitoringartificial intelligenceThesisOpen access