The sieve is the popular name for the tests that soccer clubs hold periodically when searching for young talents for their youth teams. Dozens — sometimes hundreds — of boys are grouped in a soccer field and distributed among different teams, which play against each other. Club employees watch the players and, at the end of a series of short games, separate those that stand out for some reason, such as for good ball control or because they position themselves well on the field. In general, the boys are chosen essentially based on the “trained eye” of the sieve coaches to predict future stars.
Researchers at the Center for Research in Mathematical Sciences Applied to Industry (CeMEAI), one of the 17 Research, Innovation and Dissemination Centers (RIDC) supported by FAPESP, created a virtual system, iSports, that allows them to monitor and compare the performance and characteristics of the candidates using statistical analyses. “iSports can be used to identify boys whose performance is above-average in a group in a more objective way and, thus, discover soccer talents,” explains one of the developers of the virtual tool, Francisco Louzada Neto, a professor at the Institute of Mathematical and Computer Sciences of the University of São Paulo (ICMC-USP) in São Carlos and coordinator of technology transfer for CeMEAI.
The systems consist of a proprietary platform that must be provided with data on players’ performance in different types of physical tests, including resistance and strength, and also on skills that are important when playing soccer, such as dribbling a ball around 5 cones, kicking, and passing. The researchers in São Carlos established partnerships with two soccer schools in the city, which conducted tests involving more than 100 boys. The information collected was transferred to iSports, which has an open-source module for statistical analysis to compare the different performance indices of a given boy with those of the other members of the school. Thus, after the results of the tests are input, the program indicates which player would be most skillful, or most fit. If the same group of boys is tested more than once, the system also calculates which of them has a more consistent performance. “We can compare each member of the group individually or place them on different teams and compare one team to another,” says Louzada. iSports was described in an article published in the February issue of the journal Expert Systems with Applications.
The Z strategy
The use of statistical techniques to identify sports talents is not new in Brazil, although implementing more scientific approaches in clubs and teams, especially in soccer, is never an easy task. The basis of the iSports statistical analyses, which simultaneously take into account more than one variable related to an athlete’s performance, is an old formula for identifying promising athletes that had been somewhat forgotten: the Z strategy. Developed in the late 1980s by the physician Victor Matsudo and his team at the Research Center of the Physical Aptitude Laboratory in São Caetano do Sul (Celafiscs), Z won the award for the best scientific work presented at the 1992 Cultural Olympiad, held in Barcelona. The strategy allows one to calculate the best (or worst) performance of an individual at a given task or exercise in relation to the population average for the same sex and age. The higher the Z index, the better the athlete’s performance with respect to his peers.
The logic of the approach appears to be simple and accurate. However, Matsudo had some reservations about the use of the Z strategy to evaluate high-performance athletes. “Unfortunately, in Brazil there are still initiatives that claim that they can identify a great Olympic talent by analyzing just 100 promising youths,” criticizes the physician, who now focuses on studying physical activity as a way to improve health. “The experience of the former East Germany — the greatest talent scout of all time, given the size of its population — indicates that one can find 10 super athletes in every 100,000 children evaluated.” Matsudo employed the Z strategy on 7,000 children in São Caetano. The basketball player Hortência was one of the Olympic athletes evaluated as a child by Matsudo, who identified her as talented when she was 12.
At the moment, iSports is being tested on more children abroad than in Brazil. The RIDC researchers established a partnership with Josivaldo Souza Lima, a physical education professor who works in the Advanced Research Center for Exercise Physiology in Talca, Santiago, Chile. “We are using iSports on a sample of 30,000 children and adolescents, aged 12 to 16, in public and private schools in all 13 regions of the country,” explains Lima. Using the comparisons provided by the system, the Chileans plan to filter promising athletes for several sports — not just soccer — and send them to high-performance training centers. “Through this work we hope to aid the government’s Olympic aspirations for the 2020 and 2024 games,” says Lima. According to Louzada, iSports is already ready to be used in soccer schools and can easily be adapted for talent scouting in other sports.
Center for Research in Mathematical Sciences Applied to Industry (CeMEAI) (nº 2013/07375-9); Grant Mechanism Research, Innovation and Dissemination Center (RIDC); Principal Investigator José Alberto Cuminato (ICMC-USP); Investment R$11,556,885.76 (for the entire project).
LOUZADA, F. et al. iSports: A web-oriented expert system for talent identification in soccer. Expert Systems with Applications. V. 44, p. 400-12. February 2016.