This study explores the development and analysis of a metric termed ”Manager Coefficient,” designed to quantify a football coach’s contribution to team performance. The metric was developed within the course ”Data Science Applied to Football” and aimed to provide a standardized measure of a coach’s impact on match outcomes. Data was collected from FBRef and Wikipedia, encompassing Premier League matches from the 2016-2017 to 2023-2024 seasons. The study involved the formulation of two models, the Expected Points (xP) model, to measure how favourite a team is in a match, and another one called Expected Points Overperformance (xPovr). From those two models, the Manager Coefficient is the value the feature of who the manager was in a match had in explaining the xPovr.
This value was then analyzed further as a possible metric to measure the influence a manager had in the result of a match, however, the metric’s predictive validity was found lacking, especially in predicting future trends or when applied outside the training dataset context. This study concludes that while the Manager Coefficient offers some insights within its specific data and training context, it does not provide reliable predictive power when generalized. Future research will explore alternative models and metrics, incorporating tactical and methodological compatibility between coaches and teams.