28.06.2018
Research news
University news

Talented soccer players can be identified at a young age

Researchers at the Faculty of Information Technology at University of Jyväskylä have developed a method for identifying talented soccer players. Talents were recognized among young players with 80 percent accuracy.

Based on physical and technical skill tests and psychological self-assessments done at the age of 14, players that advanced in their careers at a later age were identified. Among a group of one thousand players, seven players that in the future signed a contract with an academy abroad were recognized.

An artificial intelligence method designed for detecting anomalies, e.g., in cyber threat detection, was used. The same method could be utilized in other similar applications, such as recognizing people at risk of social exclusion or high-risk patients.

Recognizing the factors that separate thousands of players is challenging because the factors behind success are very complex. Therefore, traditional explanatory statistical analysis is not sufficient for the task. A typical problem with big data is also that longitudinal datasets are scattered and have a lot of missing values. In addition, the amount of potential elite athletes is small compared to the whole player population; thus, the problem was approached with anomaly detection, namely a one-class support vector machine. The method was first trained with data from “normal” players. When it was tested with unseen test data, it was able to recognize the players who had contracts in other countries.

The method achieved a total accuracy of 80 percent. It was sensitive enough to recognize the Finnish players in the data that have signed a contract to play abroad, and the next phase is improving the accuracy. At the moment, the method also identifies other players that have not signed contracts. In practice, the method recognized possible potential; there might be many reasons why these players have not advanced in their career.

The method can give additional information and help coaches with decision-making. In the future, the method could be developed for player career guidance for specific positions. Separating players into attackers and defenders may be useful due to the different requirements. More data is continuously being collected at Eerikkilä Sport & Outdoor Resort, and the research continues.

The research was conducted in collaboration with Sami Hyypiä Academy at the Eerikkilä Sport & Outdoor Resort as part of the project “Value from Health Data with Cognitive Computing”, funded by Business Finland.

More information:
Docent Sami Äyrämö, tel. +358 50 3255 685, sami.ayramo@jyu.fi
Doctoral Student Susanne Jauhiainen, tel. +358 40 805 3652, susanne.m.jauhiainen@jyu.fi