Getting the Boot? Predicting the Dismissal of Managers in Football

Abstract

Football club managers have a challenging and remarkably volatile job—the practice of sacking and replacing managers is widespread in the modern game. However, it is still unclear what exactly motivates managerial dismissal in clubs. More than ever, high-quality statistics are available to clubs, suggesting that dismissal decisions tend to be well informed. Likewise, supporters on social media might also influence clubs’ decisions. Here we propose machine learning models to characterize the determinants of managerial dismissals. Is social media pressure associated with managerial sacking? Yes! We fit multiple ElasticNet regularised logistic regression models using features based on the social pressure of fans on Twitter and football statistics, showing that our best model obtains a balanced prediction accuracy of 0.75.