Using Weather and Playing Surface to Predict the Occurrence of Injury in Major League Soccer Games: A Case Study

Injuries in professional soccer games are very common and can greatly impact players, teams, and leagues. The ability to predict conditions under which injuries are likely to occur would help to mitigate competitive and financial losses. This paper presents a case study in which we look at injuries during 713 Major League Soccer games spanning the 2015 and 2016 seasons. Our dataset consists of 713 regular season games, 548 of which recorded at least one injury. In total, our dataset includes information on 1,238 separate in-game injuries. In this paper, we compare the performance of nine different classifiers for predicting occurrence of injury based on local weather and playing surface. We find that Support Vector Machine (SVM) works best with this dataset, while three of the other classifiers also yield comparable performance. Further analysis shows that our results are statistically significant. We are presenting this study as a proof of concept in which we are able to confirm the efficacy of using available weather and surface data points to predict likelihood of injury and identify a path forward for future research.