Prediction model aims to predict footballers' injury risk

The model could "potentially prevent injuries in the future," Hecksteden explained in an interview, highlighting the benefits of the model. The ultimate goal is to "identify players who are at high risk of injury" and thus intervene and counteract with "personalized, medical approaches," the sports physician, who conducts research at the Medical University of Innsbruck and the University of Innsbruck and brings many years of experience in elite sports as a doctor for German national teams – for example, national teams at the Youth Olympic Games or World Games – to the table, explaining the approach and goals. The benefit is particularly great "when players return from a previous injury," Hecksteden added. The current, renewed risk of injury represents an "important basis for determining the right time to return to play."

In any case, the "identification" of players at high risk of injury and thus the risk assessment are inadequately achieved with the largely data-driven forecasting models currently used in professional football. These produce a veritable "tsunami of data" that is difficult to interpret precisely. "This includes data from the areas of athletes' workload and performance, GPS data, countless videos, or even AI skeleton models," she cited a few examples. There are reasons for this inadequate performance: "There are simply too many variables being brought into play, which means that the focus on the essential factors is lost," said Hecksteden.

"Subjective perception" important factor in model

The now accelerated model – which was presented in June in a commentary in the journal "Science and Medicine in Football" – works differently: "We deliberately reduce the complexity of the model to a few parameters." In any case, the aim is to "measure the important, actually decision-relevant aspects" and also incorporate human expert knowledge.

"This is important, for example, when it comes to player fatigue, which increases the risk of injuries," Hecksteden explained. A player "feels the fatigue" and even an "experienced coach perceives it," but it's difficult to capture with objective data, the sports physician explained. In this regard, there is "the unique characteristic of human perception," and "subjective perception" is "difficult to model," the scientist emphasized. The "coach's eye" thus supports computer-assisted processes, which will continue to be used "to optimize the models."

Another problem with the models and procedures currently in use is the small sample sizes. "We're not talking about online marketing with millions of data sets," Hecksteden said. "The sample sizes in top-level football are relatively small because we don't have millions of players at our disposal." After all, we're talking about "an elite" here, she emphasized. The creation of purely data-driven, highly complex models that are then "also applicable to new players" is therefore only possible to a very limited extent.

"Practical model" for top-level football as a goal

Now, following the commentary, in which the new model was illustrated using two real-life examples, it's time to validate the forecasting model with additional cases and thus underpin its effectiveness. "We need to take a closer look at the benefits of expert adjustments to the model," explained Hecksteden. This should ultimately result in a "technically plausible, practical model that can be applied in practice in elite football and possibly also in other elite and team sports," the researcher said.