Team sport is a sport in which opposing teams interact directly and simultaneously to accomplish a goal that frequently entails teammates facilitating the movement of a ball or similar object according to a set of rules. Team sports are a key component of the physical education of most children and adolescents, with a number of additional benefits including the development of social skills and an active lifestyle.
Many of these sports also provide a competitive outlet for individuals, which can motivate them to push themselves beyond their comfort zones and achieve their fitness goals. However, playing these types of sports can be time-consuming and often requires more effort than exercising alone. It is therefore important to balance training with other methods of exercise such as strength training and cardio, in order to maintain optimal health and performance.
The increasing prevalence of tracking systems that quantify locomotor characteristics in team sports has led to an increase in the ability to objectively describe and monitor training and match performance [1]. A primary function of this technology is the identification and reporting of external load, which can be assessed in a variety of ways, including comparison of the individual performances of athletes (e.g., between playing positions), bio-banding by age or maturation, and the use of rolling averages versus discrete 5-15 min epochs [2].
There is currently a broad range of metrics that are used in team sports to evaluate and monitor training and match performance, with the aim of improving athlete preparation, management of injury risk, and quantification of competition characteristics [3]. The selection of these metrics, however, requires critical thinking and consideration by practitioners and is often limited by the limitations of the underlying monitoring system and its associated algorithms.
The relative values of absolute total distance, acceleration and deceleration are commonly used as metrics to identify high-intensity actions in team sports. These actions, such as turnovers, cuts and close outs in basketball, or defensive shuffles and body contact in ice hockey, are largely characterized by short-term changes in velocity, but have high amplitudes. Moreover, the duration of these actions can be very different depending on the context of the game, which has important implications for physical preparation and prediction of match performance.
This issue can be resolved by utilising time-series analysis instead of discrete epochs, which eliminates the need for pre-defined time windows and enables the detection of changes in the mean and variance of a metric. In this way, the characteristics of a metric can be explored without being influenced by its context and may allow for more targeted training drills that better prepare athletes for future match demands. In addition, this approach allows for the detection of features from raw GPS or LPS traces that are associated with peak match intensity. This can be done by visualising a distribution of features and their corresponding frequency. This reveals the moments when players reach the most intense speed.