Team sport involves a group of athletes who compete as teams to achieve team goals. It is a dynamic, high-intensity sport requiring flexibility, endurance, strength (to catch and hold teammates), and concentration. It also requires an extreme reliance on one’s teammates and is often performed in crowds.
Team sports involve a complex, multi-component training program that includes specific characteristics (e.g., agility, plyometrics, strength and power) that vary with the specific roles players play in each sport. These characteristics are required to perform in the game and may differ between periods of a training session or during each drill.
In the context of team sport, practitioners plan athlete external load in order to maximise performance, reduce injury risk and meet team goals [1, 10]. The process of developing a training adaptation or response from drill to macro-cycle level, and the subsequent evaluation of these adaptations is a key element of a performance staff’s role. Practitioners use tracking systems to collect data, where derived metrics provide the information they need to evaluate these adaptations.
The selection of appropriate tracking systems and derived metrics to profile athlete physical characteristics during training and matches is an important consideration for practitioners. The type and level of technology used will vary according to the sport. However, a critical assessment of the validity and accuracy of each system will be necessary in determining its suitability for the sport.
Tracking systems are a great tool to capture athlete external load and thereby assist practitioners in planning training adaptations and monitoring performance during a match. This type of analysis can be performed in real time during live events, or by using retrospective data to analyse individual sessions.
An increasing number of tracking systems are now available and offer a variety of derived metrics for sports. Despite the availability of tracking technology, there is a lack of consensus on the most appropriate metrics to characterise an athlete’s physical requirements across different phases of a training session or match. This can lead to a significant challenge for practitioners.
Several approaches have been proposed for the metric selection process, including fitting Gaussian curves to instantaneous velocity data or combining features with a moving average. The latter approach has the potential to overcome some of the problems associated with relying on pre-defined speed thresholds, given that the movement profiles of team sports are highly dynamic and rapid.
Another approach, based on frequency domain analysis and time series segmentation has been applied to Australian football to detect how physical output changes during a match, as a function of time. Using this approach, it has been shown that peak movements in a match occur around the time of skill involvement, and that these demands are significantly greater in away games than at home.
Similarly, it has been shown that skilled output is largely unrelated to the aggregate physical characteristics of an athlete, when analysed via linear mixed models or conditional inference trees. Despite the presence of a relationship between the two, the explanatory power and accuracy of these models are limited and suggest that subtle changes in athlete physical and skilled output do not necessarily appear as peak characteristics.