ABSTRACT
Until recent years, sports analysts have typically used what are now seen as rudimentary statistical methods primarily to process strongly labeled data. These data types are traditionally difficult to obtain. However, technological developments are now occurring which enable data to be automatically generated directly from video imagery, thus removing humans-in-the-loop and automatically producing as human-interpretable performance statistics. As part of the 2020 Olympics in Tokyo, Alibaba/Intel's jointly developed 3D Athlete Tracking application will be used to create 3D representations of athletes in action from which coaches can extract real-time data. Alibaba and Intel provide the two vital components – video quality and processing power – to create the final product, which is a digital model of the performance. Another significant development, in team sports especially, is the increased understanding of the use of space that machine learning can provide. All sports are now selling off-the-peg versions of bespoke artificial intelligence technology initially developed for the elite market.
