Development and validity of the Playermaker ball touch classification

Kicking and ball interactions are a key component of some team sports such as football (soccer), Australian Rules Football (AFL), American Football (NFL) and both Rugby Union and League (Lees et al., 2010). The interaction with the ball is dependent on the sport but is representative of a method to score more goals/points than an opponent. In football (soccer), the technical and tactical statistics of kicking are regularly used by to measure performance indices including a breakdown of the number of passes, shots and crossing performed by one team against another (Barnes et al., 2014; Liu et al., 2015). These actions are currently used by coaches, performance analysts and scouting/recruitment departments, in order to assess a team and individual players performance within a given situation (Yi et al., 2019). These statistics are now openly available via media streams which allow the fans to engage with these statistics during competition. Semi-automated and manual coding methods are currently used in order to extract this data from match play and training using systems such as Opta, Stats, StatDNA, Hudl and many more (Liu et al., 2013). These systems require a number of operators in order to ‘tag’ the kicking events within sports and can be subjected to potential human error during these methods of tagging (Liu et al., 2016). When analyzing a professional football match in Spain (Real Madrid vs. Valencia), Liu and colleagues (2013), found that there was a very good agreement between operators (0.86-0.94). However, 4 operators were required for this method during one match in order to provide the data within 24 hours to the clubs, using multiple HD cameras situated around the stadium. The expense of these systems means that data collection tends to be restricted to within match situations and limited technical and tactical data is collected within a training environment as a result (Rein & Memmert, 2016).