A machine learning and data-driven assessment model for characterising and predicting team success in Championship football

Football players waiting to play

This project will develop novel metrics to characterise a team’s style of play within the EFL Championship. The metrics will be derived from in-possession and out-of-possession team statistics as well as individual performance indicators, and will be refined through various data reduction techniques before being combined with Dixon-Coles models to predict individual match and league position outcomes. The project will also apply these performance indicators to legacy data to retrospectively establish their efficacy in match prediction, whilst also investigating whether the performance indicators can be reduced in order to be more actionable to practitioners at the individual player level, and also to provide a scientific basis on which to potentially adjust future team structure, playing formation, and strategy. 

This project could also help football practitioners understand other physical consequences of playing patterns and the ‘effort-level’ imposed on individual players – a potentially exciting new avenue that will help redefine post-game recovery for players as a function of style of play.