How AI is empowering actors in the football industry

The world we live in is becoming more complex. Information technology makes it possible to generate an unprecedented amount of data. The challenge today is to know how to benefit from it and the actors of the football industry have understood that.

The use of mathematics in sport has been around for almost twenty years. For example, Bennett Miller’s excellent movie "Moneyball", released in 2011, traces the story of Billy Beane, general manager of Oakland Athletics, which relies on statistical theories and revolutionizes the practice of one of the the most popular sports in the world. Recent technological advances, like storage and calculation capabilities, now allow us to go even further in the use of data and its applications. 

There are many applications of artificial intelligence in football. Before discussing the specific uses of this field, let us recall that a professional footballclub operates in a similar way to a company in some regards. Especially in terms of image and marketing. These aspects should not be overlooked when we know that a majority of a football club’s income is generated by the sale of jerseys, tickets and sponsorship.

Data science has already proven itself in many applications around marketing [1]. Data analysis improves customer knowledge to tailor offers. Clustering methods, predictive modeling or factor analysis are all techniques that are used to accomplish these tasks. It is also possible to improve the customer experience with the use of recommendation algorithms on websites. Finally, there are value-added applications such as predicting which customers have high potential or predicting sales and the impact of a marketing campaign. The techniques used in this case are predictive modeling or time series analysis. All these methods are available to clubs, who have a strong interest in using them if they want to increase their audience’s engagement and their revenues.

We are now in July 2022, France is defending its world championship title in Qatar against Belgium. In the 75th minute France is leading 1-0 and several strategies are on the table for the coach. Play it safe and make a defensive change? Add an attacker and try everything to prevent the Belgian team from coming back? Or change the play-system? He then leans over to his computer wondering, "What changes do I need to make to have the best chance of winning? ". The computer responds, "With a defensive strategy you have a 64% chance of winning". Of course, France would not be the world champion in regards of “beautiful” game, but would be able to claim another championship star.

Anticipation, of course. We are not talking about France defeating again Belgium, but about the use of artificial intelligence as a decision-making aid during a football game, or any other type of team sport. This might seem like a fantasy today, but let's analyze the possible advances in this area.

First, recent improvements in speech-to-text, a speech-recognition technology, make it possible to transform an oral speech into text in an automated way. In addition, natural language processing allows a computer program to understand human language as it is spoken. It is now possible for a machine to understand and answer oral questions. It could therefore calculate several live forecast scenarios and propose a tactical strategy or changes to be made. Indeed, by analyzing beforehand an important history of games, it is possible to statistically simulate which scenario would be the most relevant at a given moment. The coach could also interact with the model and analyze the influence of the game’s variables. It could for example make virtual changes and analyze how it would affect the game. It is on the basis of this application that the Paris Saint-Germain football club and Ecole Polytechnique are collaborating to develop the best live prediction algorithm [2].

Some football clubs, such as Liverpool FC, which has partnered with the French start-up SkillCorner [3], are using powerful tools that can automatically collect real-time data from video recordings of games or training sessions. This technology is able to monitor players, the ball and their performance, which is known as "tracking". In addition, the algorithm can also highlight key factors such as abnormal ball possession or a frequent area of ​​ball loss. This opens up new opportunities for data-driven decisions.

For coaches and players, who will be able to access many additional statistics. The Scottish club Glasgow Rangers, for example, is working with the American platform Hudl [4], which records and analyzes all the games in the league. This provides a coach with the opportunity to know his team better but also his opponents’ team. In addition to accessing classic statistics, namely the number of shots, passes, games won, etc., it is possible to enhance artificial intelligence to preselect game phases. For example, when analyzing a game, the coach may choose to observe his attacker’s shots, or the ball losses causing a dangerous action. This would allow him to quickly obtain new data and identify the strengths and weaknesses of his team, his opponent’s and develop strategies accordingly. This tool is not only for the management, players can also use it to analyze their performance, take a step back and identify areas of improvement. The possibilities are almost endless because the coach as well as the players, can have at their disposal many complementary statistics to those that exist today. In addition, the time saving is considerable during the preselection of the phases of interest of a game.

Other teams use connected objects, such as clothing or wearable devices. Thus, Real Madrid and the French Football Federation are working with the company Catapult Sports [5], which has developed a technology that can be put in the back pocket of the player's jersey. Manchester United and Juventus are working with STATSports [6], one of the market leaders in connected clothing. These include the ability to retrieve data such as speed or distance traveled, which can be analyzed later.

For the detection of new talents, because improvements and lower costs related to storage and speed of execution of algorithms, this technology can be extended to all games (not only European). We can therefore use it to carry out scouting (talent recruitment). Indeed, we can already create a large database containing players with their classic features. Today, with artificial intelligence capable of analyzing matches, we could enrich this database with very specific additional statistics. These are no longer limited to the number of goals or successful passes, but to more complex metrics that can take into account multiple factors. Our perceptual limitations prevented us from recording these metrics and evaluating them accurately, this is no longer the case with artificial intelligence. Thanks to this enhancement, we can train a model capable of detecting aberrant behaviors. It may output a pejorative or meliorative connotation, the latter will be to target the players with great performance and potential. The Liverpool FC club who have been using automatic talent detection for some time now, and in more depth in recent seasons, have made a series of interesting recruitments. This may explain their victory in the champions league during the 2018/2019 campaign. In addition to talent detection, the enhanced database can help us estimate the players value on the market. This can be useful during negotiations, whether buying or selling a player.

For television, a tool that collects data automatically during games, without human intervention, can revolutionize the live broadcasting and influence the way people watch sport. Channels will be able to publish live information about the players or teams performance, or show key statistics that can not be detected by a human in real time. Moreover, by using historization it would be possible to compare the performances and the matches with previous ones. This enhances the customer's experience in front of his TV channel and increases his engagement. By watching the game, the viewer may be able to access live statistics and analyze the match more easily. For some years already it is common to see them displayed on screen. However, artificial intelligence makes it possible to collect new data automatically and process it more efficiently. Some channels such as Canal +, which innovates a lot in retransmission, are already using new computing devices to enhance viewers' experience. Other media such as L’equipe and TF1 use an intermediary like the English company Opta Sports [7].

For sports bettors, the benefit is similar to that of the TV channels. The Paris websites want to increase the loyalty of their customers. Winamax has also partnered with the French start-up SkillCorner [3] to benefit from a tool that provides them with real-time key information about the game. The customer could then have valuable help with his bet. As with artificial intelligence, the more quality data it has at hand, the better the prediction will be. The bettor will therefore be more loyal to a site that provides him with key information and help in his prognosis. In addition, sports betting sites could consequently better adjust their sport ratings, because these are set by bookmakers but also rely on mathematical equations. Today we can bet on a lot of games, not only on the victory of our favorite team, but also on the exact score, who will score the first goal, at what minute, etc. Thanks to artificial intelligence, betting sites are then able to adjust a large number of sport ratings automatically.

For health, the impact and predictive power of the algorithms are already well known. Physical preparation for high-performance athletes is essential. Sport teams have always invested a lot in players’ physical and mental preparation. Artificial intelligence can analyze various health parameters, monitor players’ movements, evaluate their shape and even detect signs of fatigue or stress that could lead to injury. VALD [8] is already developing solutions to assess the body’s resistance, imbalance or asymmetry. Another prevention system, proposed by Kitman Labs [9], also reduces the risk of injury for athletes by identifying warning signs. This software can therefore greatly assist the medical teams to maintain the players at their best physical level, avoid accidents and thus retain their most valuable assets for long seasons and at the best times.

We have seen the benefits of using artificial intelligence in football. This can of course be applied in the same way to many other sports. Technology is constantly changing our society and sport is no exception. In the United States, the NBA and the NFL are already well ahead. In particular, since 2013, the NBA has installed tracking systems in each of its 29 stadiums which allows to collect an impressive amount of data, that are used by all clubs. It also launched a challenge to predict the scores of games [10]. Many NFL football players, for example, are equipped with sensors that send 25 signals per second monitoring their position and speed. The advantages of these inputs are obvious but this requires of course initial investments. Also motor sports, and more particularly F1, are starting to use artificial intelligence. Indeed, thanks to several sensors placed on the car, it is possible to analyze all the data in real time and detect anomalies. Thus the race teams are able to improve the car’s performance before and during the race. This is the most popular application today but we can easily imagine a tool that would analyze the races and which help determine the optimal strategy to change tires or pit stops.

Once data is being collected and used on a large scale, it is legitimate to question what the legal framework is. Should we allow the collection of any type of information, from our opponents? For which purposes? In some regards, can it be considered possibly as espionage? Industrial doping or unfair advantage? We can also discuss the availability of these technologies, infrastructures and data to everyone. For example, for the French Championship, unlike the Spanish, English or German, the lack of sharing of tracking data is a limitation. What seems inevitable is the evolution of organizations to better understand the benefits of artificial intelligence. This requires the adoption of new skills around data science as well as new technological partnerships.

We shall conclude by clarifying two key ideas. Statistics and other indicators are becoming omnipresent. However, we must be vigilant when using them because a statistic alone is not necessarily relevant and in the context of sport it does not necessarily summarize a performance. It is important to consider the context and to carry out a “big picture” perspective. Is it possible to replace intuition and feeling? Zidane did not have the best statistics yet was he not a great player? This leads to our second point: artificial intelligence is not "intelligent" in the human sense, it should rather be taken in the anglo saxon sense of the term, "information". AI is capable of analyzing data and processing large volume very fast. It is a great tool that can have a positive impact in helping you make decisions and / or avoid time-consuming tasks. However, part of the sport is about feelings, talents, humans. Is it still possible to model that with algorithms? How to find a balance between the beauty of sport and technology? Augmented coaches, the ball is in your court!

[1] https://towardsdatascience.com/20-practical-ways-to-implement-data-scie…
[2] https://www.lebigdata.fr/psg-intelligence-artificielle
[3] https://www.skillcorner.com/
[4] https://www.hudl.com/en_gb/solutions/professional
[5] https://www.catapultsports.com/
[6] https://statsports.com/
[7] https://www.optasports.com/sectors/media/
[8] https://www.valdperformance.com/
[9] https://www.kitmanlabs.com/
[10] https://www.nbadatachallenge.com/

Contact : david.martineau@sia-partners.com