A UK professional rugby club.
Systematically scouting new players for the club to recruit is incredibly time-consuming, despite the accessibility of a great deal of performance statistics on individual players. The selections clubs make are also subject to human habits and preferences that may overlook the value of some less well-known players. The club asked us to develop a machine learning model that would help to increase the efficiency and effectiveness of their player recruitment process.
We ingested the rugby player performance data from OPTA, which was formatted as a series of Excel spreadsheets, into Faculty Platform. We then developed an application for the use of the rugby club staff.
The application enables a user to input the name of any professional player in the game and instantly return a list of other players using sophisticated clustering algorithms to identify players with similar characteristics. These deliver more accuracy than more traditional statistical techniques such as linear regression.
Sample output from the player comparison application.
In addition to increasing the speed at which candidates were identified, when testing the model, the club evaluated the player suggestions to be at least as good as or better than their current methods 100% of the time. This means that the club is now able to identify and bid to recruit undervalued players whom they would otherwise have passed over.
While the story of the baseball team Oakland Athletic has popularised the use of statistics to help sports teams in selecting players, this is the first time to our knowledge that data science and machine learning techniques have been used in professional rugby.
This application is now being intensively used as a recruitment tool for the club to identify and bid on players, and has been progressively enhanced a number of times to provide additional capabilities.