Researchers from The California Institute of Technology (Caltech) along with Disney, and STATS, a supplier of sports data have collaborated to teach a computer to recognise soccer games the same way a fan would. The neural network was trained to recognise the roles of individual players, when they switch positions and the formations being used. The algorithm can also be used to imitate player behavior, and guess how a player will respond to a particular formation.
Yisong Yue, who collaborated on the study with the lead author says, ""We're training the algorithm to understand soccer at the same level that a fan would. It's not just mindlessly watching faceless players move across a field; it's watching strikers and right midfielders and forwards arrange themselves in specific formations." The machine was able to determine the role of a player in a team based on their movements across the field, even though the raw data had not been fed in.
The findings of the research has been presented at the International Conference on Machine Learning in Sydney, Australia. The implications of the study extend beyond the field of sports. The same underlying computational technologies can be used to control teams of robots in emergency response situations, plan platoons of autonomous vehicles, and model the behaviors of groups of animals.