I recently uploaded my thesis on arXiv. My thesis, called Predictive Modelling of Football Injuries, was a collaboration with Tottenham Hotspur FC and dealt with the problem of predicting injuries in football.
Sports analytics is a relatively new field and there are very few people who are researching the area of injury prediction. My thesis makes three main contributions to this field:
- It studies the use of machine learning for predicting the recovery time after an injury has taken place.
- It studies the problem of predicting the first injury of the season for a player, based on exposure records. This is done through a custom method developed for this project that consists of a Gaussian Process model with a Dynamic Time Warping kernel.
- It provides a way to predict injuries based on GPS measurements gathered during training sessions. This was
- done through Supervised PCA.
The most common football injuries
I believe my thesis will be of great interest both to sports scientists and data scientists who want to read more about sports analytics. Some of the lessons learned are valuable for other sports as well and some of the models used could be copied over to different instances in a pretty straightforward manner. i also have a section regarding data standards, data strategy and data culture. My work in this project convinced me that the “soft parts” of data science are as important as the “hard” parts. It is just not possible to do quality work if the soft parts, like data collection, have not been taken care of properly.
For those of you who are interested to study more about the subject of I have also released a video training series with Experfy (also make sure you check out my other courses). If you get in touch I can provide you with a discount coupon. Feel free to e-mail me if you have any questions or comments.