The revolutionary method of Barça to try to predict injuries

BarcelonaInjuries are the main headache for all football coaches. Some are unpredictable: a bad move, a fall... situations difficult to control. Others, especially muscular ones, can be prevented if several variables are taken into account, such as accumulated distance covered or intensity. In this regard, the Barça club has been working for five years on the idea of designing a model that allows for anticipating injuries in women's football.

The latest artificial intelligence model presented by Barça Innovation Hub was published in the scientific journal NPJ Digital Medicine and was carried out in collaboration with the University of Bonn, the Barcelona Global Health Institute (ISGlobal), and Made of Genes. "We are approaching models that are between 70% and 80% accurate, which is a lot scientifically, but not enough for a coach," assures Gil Rodas, Barça's doctor and one of the researchers, to ARA.

Although algorithms are becoming more precise, the latest model has been temporarily tested in a single season, but its impact still needs to be assessed and recommendations applied. “We are developing algorithms, incorporating them, teaching the system, but until we have a good prediction range, I cannot ask the coach not to play a player in the next game because she will get injured,” points out the doctor.

More than 30% of injuries in the world of football, both male and female, are muscular or tendon-related, but one of the most concerning is the cruciate ligament injury, due to its severity. “It can be up to five times more frequent in women than in men, and it has to do with the menstrual cycle and training methods. Right now, predictions for this type of injury are not so solid; we need to consider more variables,” confesses Rodas.

A more precise and individualized algorithm

Starting from a dataset collected between 2019 and 2023 from Barça's first women's team, a series of variables are incorporated into the model to estimate when an injury may occur. The peculiarity of this new study is that it allows differentiation between training and match loads and takes into account accumulated load. “We measure load like a car measures the kilometers it has traveled and the fuel it has consumed. Load is the distance a player covers during a match, but also the speed she uses,” he explains.

“The GPS devices that footballers wear measure more than 20 and 30 variables. These algorithms allow you to see if it's better to include sprints at 30 kilometers per hour or the distance of a match. We take into account various variables, and the beauty of the research is trying out different combinations,” describes Rodas, who opts to keep the recipe a secret and not reveal which variables they are currently introducing. “It's copyright,” he adds.

But not only the external load —the so-called physical load— is taken into account, but the internal load is also incorporated into the model, that is, how they have slept, whether they have pain or not, how they feel, etc., variables that are introduced based on what the players themselves explain and convey.

Another of the factors that this new algorithm takes into account is the importance of each match, since a Champions League tie does not have the same demands as a league duel against a team from the lower part of the standings. “How do you value this? It’s not easy and there can be a lot of criticism,” asks the doctor, who emphasizes the importance of working in the long term.

On the other hand, one of the variables not taken into account in this model is the menstrual cycle. “In a study we did together with Dr. Eva Ferrer, we observed that half of the female footballers did not ovulate. This is a problem because in each algorithm you have to input the specific menstrual cycle of each player for it to work well, and this makes it almost personalized,” she confesses. All of this leads to the construction of a “digital avatar” for each of the female footballers. “It will be the only way to anticipate each one's injuries. This is called personalized medicine,” points out Rodas.

In the case of the male players, Barça is working internally, but is calling for more data to increase accuracy. “At the hospital, I can go online and get data on liver cancer patients from the University of Chicago. In football, we are working very specifically with our players. We should have data from the footballers of Chelsea, Madrid, Atlético de Madrid… I am trying to create within the world of football an open space of anonymized data to improve our algorithms, first generic and then personalized to each club,” Rodas claims.