March 10, 2022

Technology & innovation

Artificial Intelligence Applied to Sports

By Barça Innovation Hub.

The immediate implementation of Artificial Intelligence (AI) in sport has been directed most of all to the quality increase in decisions taken by coaches or all the people involved in a sports strategy.  There are tools that allow you to predict from opponents’ behavior to fatigue. The possibility to do personalized trainings that range from planning the loads to improving specific technical skills has also been developed.

In the recruitment market, data management is starting a revolution, especially in clubs where there is no need of immediate performance incorporations but need to bring reinforcements that adapt to specific needs.

In broadcasts of some sports, stand-alone cameras can tell key play moments to edit videos and provide statistics in real time which are related to the direction of the game.   Those numbers can later be translated into written reports about what happened in the game and that could be perfectly read. Elite competition will always need a human mind to interpret it, but the revolution this brings in mid and popular levels of sport has not been seen before.  On the other side, for referees of certain disciplines, such as gymnastics, it will be useful to assess an exercise with greater accuracy and mark it with higher precision.

Speedgate, the first sport created by AI

An implementation that has already been started is AI designing games or new sports. The outstanding example was Speedgate. A neural network driven by GPUs Nvidia Tesla which created the rules of a new method after processing basic regulations and concepts of 400 sports. However, the result was more than a thousand sports, some of them almost surreal, and a human mind criterion was needed to choose the easiest ones to put into practice. Speedgate was the chosen one. It is a kind of rugby in which possessions last only 3 seconds, and you have to score first in a center gate, and then in the opponent’s goal.

The most important of all this is that the IA development has reached a stage in which the scenarios proposed change the paradigm.  It is not just about a tool to assist specific sports or the creation of new game proposals, it is about AI practising the sport in a rather autonomous way. 

Will robots play sports?

In the half of the past decade, an article published in Sports, Ethics and Philosophy wondered about this possibility. Its title: “Will robots ever play sports?” where researchers analyzed whether in the future the creation of robots with human athletes’ behavior in a track or play field would be possible. At least there are precedents.

In 1956, the German inventor Josef Pachta designed inflatable cylinders which pretended to be rivals. The idea was the inspiration for the 1997 Japan RoboCup with 5000 spectators with the aim of having created androids able to beat a human team in the World Cup by 2050. Even though experts consider that walking with stability is complicated business for a humanoid robot, and coordination and agility in football are still far away, the initiative moves forward. Something similar happens in wrestling with the Bar-Cohen project of building an arm driven by artificial muscles which could defeat a human being in wrestling in the future.

The problem, however, would be the same AI faces in any other field. As research professor in cognitive science, Margaret A. Boden, documented: a human brain does not only base on memory and analytical abilities, qualities a machine can replicate, but on feelings. To explain it, she gave information about a patient who, after having a brain tumor removed, lost his ability to have feelings related to suffering and joy. Without emotions, without being able to feel joy and pain, his reasoning was worse than before the surgery when he was a successful businessman. This proves that feelings are key in intelligence.

In the sport case, the human factor that technology cannot reproduce is the vulnerable nature of the human being and its corresponding desire for self-improvement. These qualities are attached to sport and where its appeal lives. A robot can improve itself, in fact, that is what AI is about, but regarding the public, there will be little interest in seeing its development if its suffering is not perceived because it wouldn’t exist. When a robot athlete which is invulnerable and able to have amazing achievements exhibits itself, it won’t take part in sports, but in a kind of circus.  Achievement is not the only charm sport has, the human context in which it takes place also plays a role.

Human against machine

However, if there is one discipline that has allowed an immediate AI incorporation, it is chess. Kasparov’s matches versus Deep Blue and Deeper Blue of IBM, even though controversial, made history. Currently, machines are already better than human beings without room for discussion, and they are breaking the limits. Since 1974, the World Computer Chess Championship has existed. Its evolution reached such a point in which supercomputers and clusters were forbidden. Maybe, in this type of games, the elite would be DeepMind AI (Google), AlphaZero, which before the pandemic defeated Stockfish in a game of one thousand matches. Even more enriching was the case of using AI in the Chinese strategy game Go, which was supposed to be more difficult than chess for the computer. However, AlphaGo made moves that looked absurd at the beginning, and in the end, they were winning ones. Now they are taken into account/considered.

In Formula1, last year the first autonomous race was held in the Indianapolis circuit: the Indy Autonomous Challenge Powered by Cisco. Nine universities competed for a prize of a million dollars with real scale prototypes in a real circuit. The victory was for TUM Autonomous Motorsport from the Technical University of Munich. The race was fought over the fastest lap, and there was only one obstacle to overcome. The ‘grand prix’ was followed by twenty thousand spectators on streaming.

At a smaller scale, there have also been effective progress. For example, a table soccer driven by AI designed by the German professor Hans Michael Krause. Even though automatic table soccer already exists, it has machine learning. A camera films the playing area from above and transmits the images to a computer that analyses the movements as right or wrong with the aim of scoring goals. What it needs is a human being playing with the AI for it to learn. Once again, this puts the human being in the role of a coach of an autonomous entity one can compete against.

This side is important if we take cryptography technology into account. Digital horses races are already being held. It is a system which allows the user to create a race horse in NFT or non-fungible token format. The peculiarity is that each horse has a unique behaviour, like a DNA based on an algorithm. They could be raised in virtual stables, and they transmit the algorithm to the descendants, and it is planned that fatigue has an impact on the race results. High amounts of money have been paid for some of them and their races are transmitted on Twitch. It has been said that this could be the first digital sport of the Metaverse, but for now it is closer to eSports than to real autonomous entities who practice the sport. This possibility was clearly seen in the congress Sports Tomorrow 2020 of Barça Innovation Hub.

Bruno Dagnino and Juan Carlos Nuñez presented a study that researched the possibility of simulating a football game using AI.  The project was done through Google Research Football Environment in Kaggle, a Manchester City initiative.  The possibilities were simple, they allow you to control only one player using 19 actions. In this way, AI was trained to end up reflecting the best real play. The ultimate aim would be of great help in the real sport. Simulating matches under certain circumstances to make decisions. A way to try tactics or see how they develop, or the possible scenarios when confronting one player with another.  Like digital twins which are already being used in the industry, and they are being developed for urban planning in municipal management, they will be useful to try tactics in the virtual world, and then apply them to the real one and vice versa.

Artificial sports in the metaverse

Considering all the factors and advances, in this research lies the possibility of really going further in the current paradigm, and it could outline the evolution of one of the sports competitions facets in the future. All this development could be summarized as training bots, which could confront each other and test the human mind.  The improvement challenge would be human, mathematical but with virtual players. Something that, as seen with the cryptographic horses, won’t stop making them unique, buying and selling them, and even becoming celebrities.

In fact, there exist AI competitions of video games like Fantasy Football, the Bot Bowl tournament.  We would be talking of a sport, with the challenge it means for human beings, but artificial, virtually practiced by artificial entities, which would be what differentiates them from eSports. An artificial sport that could find all its potential with the arrival of the Metaverse.   AI experts have repeatedly explained that human creativity is still unbeatable, that is why the potential offered by the chance of creating autonomous bots that compete in sports in a virtual environment will always have human strategists behind. This time maths and programming with AI techniques should be added to the traditional tactical knowledge.



Will robots ever play sports?

Exploring the application of soccer mathematical models to game generation on a simulated environment

Human against machine: The two go head-to-head in table soccer!

Applications of Artificial Intelligence in the Game of Football: The Global Perspective





Do you have any questions about Barça Universitas?

  • Startup
  • Research Center
  • Corporate




The Form has been sent successfully.




The Form has been sent successfully.




The Form has been sent successfully.