In football, injuries are a key concern that limits players’ ability to play, resulting in both performance and financial consequences for clubs. In this project, we develop Machine Learning (ML) models for ranking football players based on their risk of injury. This project is completed in collaboration with the Danish football club Aalborg Boldklub (AaB), using data collected during training and match sessions. We fuse four datasets into a single dataset consisting of 4, 350 match and training sessions, with 89 of these sessions containing an injury. We create a ML model that exclusively utilizes handcrafted features from domain knowledge, a ML model that relies exclusively on learned features, a ML model that combines handcrafted and learned features, and a ML model that utilizes learned representations based on player ID classification to learn a player’s risk of injury. For handling the dataset imbalance during training, we utilize Cost-sensitive Learning, combined with binary cross entropy as the loss function. We are able to estimate the player’s risk of injury to some extent, providing a recommendation tool for medical staff. The best performing model exclusively utilizes handcrafted features with a precision @ k of 56.66% ± 9.08 using k = 5, with a Discounted Cumulative Gain score of 0.90 ± 0.08.