The rise of engineering and data analysis in sport
By Eusebio Angle (ESI teacher and BM Pozuelo coach)
In recent years, Data Analysis and Artificial Intelligence (AI) have marked a before and after in the sports field. To what extent do teams that invest in data engineering have an advantage over those that do not? There are more and more works that show that AI can facilitate in many high-level sports an improvement in individual or collective performance, the monitoring of players to achieve promising signings and all kinds of advantages that optimize resources and make said teams more competitive.
My interest in bringing AI closer to sport began a few years ago, when, following the advice of my thesis director Ricardo García Ródenas and several colleagues, I decided to change my line of research and try to apply the knowledge acquired in my training to sport. In this way I managed to align my work as a researcher with my passion as a national handball coach, in which I have been training for 15 years at the Pozuelo de Calatrava Handball Club.
ESI professor Francisco Pascual Romero joined this path from the outset and, shortly after, Julio Alberto López (professor of the Department of Technologies and Information Systems at the Almadén School of Mining and Industrial Engineering). In our studies, we soon became aware of the boom that the figure of the Sports Analyst was having in all sports. These sports analysts, in most of the investigations, are enriched by three areas in which our School is powerful (Fig.1): i) Machine Learning, i) Big Data and iii) Advanced Statistical Analysis.
In high-level sport, more and more resources are being devoted to harnessing the benefits of AI. There are sports that have given much value to this field, especially American sports. Some examples are basketball (NBA), ice hockey (NHL) or baseball, where deep data analysis for performance improvement is currently common. Even these analyzes are applied to athletes at an early age for the detection and recruitment of talent.
On the other hand, football is no exception and the figure of the sports data engineer is on the rise, as is the case with Sevilla FC, which has a team of engineers who apply Big Data to track players. The club itself, along with FC Bengaluru UTD, has recently launched a hackathon for performance evaluation based on objective data.
In the last two years, as a result of our scientific work in the field of sports analytics, members of the MAT and SMILE groups, have actively participated in congresses and published articles whose main theme is focused on the application of AI to high-performance handball teams [1]-[5], in search of objective evaluation of performance of players in real matches, all through the application of methodologies fuzzy and optimization algorithms. In these works, the main contribution is the objective evaluation of the players in matches and championships based on all their actions in the matches (Fig. 2). In addition, it allows a characterization of the players depending on their demarcation and knowing what qualities each player should have in his position to have an outstanding performance in a championship.
In the same line of work, the group and the University of Castilla-La Mancha have signed research agreements with several high-performance clubs such as the Super Amara Bera Bera, recently champion of the Women's Division of Honor (Iberdrola Warrior League), or the Veszprem from Hungary, a men's team that was recently a finalist in the Final Four for the Champions of handball These agreements have given us direct access to real data with high performance teams and the possibility of carrying out an advanced study to improve the performance of their own players individually and collectively, as well as a better analysis of their rivals for a good preparation of the approaches of the matches in said competitions.
In Fig 3. the values or weights of each positive and negative action are represented. The positive actions of each player are the ones that bring a team closer to victory, while the negative ones bring it closer to defeat. In addition, these weights vary depending on the player's demarcation, that is, they are more or less important. The accumulation of all the actions of a player in a match will offer us the valuation of him or scoring at the match.
These data analyzes offer an objective view of each player both in an independent match and in an entire season. In my particular case, the analyzes have been of great value during the past season, especially for the detection of drops in sports performance due to physical or mental problems. This analysis has allowed the detection of problems that were resolved in time. To the above, it must be added that sports performance can also be obtained against one's own or the rival's defensive variants. In short, a lot of valuable information can be generated to improve decision-making. Players and coaching staff have adapted well to this feedback provided by the analysis of data in each game to improve the performance of the team.
Coincidence or not, the Pozuelo Handball Club has had its best season at a sporting level with a qualification for the promotion phase to the Iberdrola Guerreras League and the Spanish championship with the women's youth team.
In conclusion, AI has provided significant advances in sports through predictions with a high degree of reliability, objective evaluations at the individual and collective level, as well as improvements in performance. Particularly, in handball there was little progress in terms of data analysis and much remains to be done, but the first steps have been taken, showing that these methods can be very useful. The future will increasingly bring AI closer to sport to have an advantage by providing a better service at all scales: athletes, coaches, sports fans, sports directors and/or sports analysts.
[1] Romero, Francisco P., Angulo, E., Serrano-Guerrera, J. and Olivas, José A.
