
With the development of Artificial Intelligence, new possibilities have come up for a lot of different areas. Soccer might not be the most prominent but thinking about potential use and the development of artificial intelligence in the last decade, connecting it with soccer is not very far.
For a long time, there was a cultural hesitation regarding the integration of data science into soccer and an overdependence on gut instincts (Tuyls et al. 3). The origins of soccer analytics are typically traced to Charles Reep, who began recording observations, because of a team’s scoring attempts patterns (Luzum and Model). Later, Reep became “soccer’s first known analytics employee” at Brentford, “which avoided relegation after winning 13 out of 14 matches and doubling their goals scored “(Luzum and Model). With technical development, the human being recording all the passes and shots was replaced by a technological system (Luzum and Model). In 1999 a business named Opta (founded in 1996) partnered with the English Premier League to “provide data insight from matches” (Luzum and Model). A year later, the information also became accessible to individual players (Luzum and Model). A rival company from Opta, named Prozone, started to collect its own data, while “the actions and metrics tracked remained highly secretive” (Luzum and Model).
The development of pixel-tracking cameras helped to track a player’s movement (Luzum and Model). The GPS (global positioning system), a tool to track players, includes basic information as: distance covered, speeds from the distances, accelerations, decelerations and the heart rate (FourFourTwo 16:10). When more soccer clubs saw a need for those services from analytical companies, the companies realized the growing popularity of data retrieval. By the end of the 2010s decade, analytics had become usual for the top European leagues (Luzum and Model). But even if analytics was more popular now, not everyone was convinced that it would help the team, like an Orlando City Soccer Club spokesperson said: “nothing can replace actually meeting a player, understanding his character on and off the field, and seeing if the chemistry works well with the organization.” (Luzum and Model).
Major breakthroughs in the past decade allow image classification, video action recognition and pose estimation (Tuyls et al. 6). Next to tracking and counting passes, the “availability of new data types enables a descriptive analysis and a prescriptive one”, resulting in an increased pool of information (Tuyls et al. 6). Today there are soccer departments that have to contend with an avalanche of information, originally collected for the fans (FourFourTwo 3:30). The most recent big change in soccer from artificial intelligence is the Video Assistant Referee (VAR), who watches the game along and assists the referee in difficult situations or mistakes of the referee in important decisions.
Compared to baseball or basketball, which are both high scoring games, soccer is hard to analyze and filter the important statistics. Baseball and American football for example are both segmented, which makes it a lot easier to do the analysis (FourFourTwo 12:21). In addition, most statistics are collected if the action leads to a goal afterwards, therefore in soccer there are essential actions, without an impact on a goal (Decroos et al. 1). One example could be a player running forward, making space and taking a defender away, which would allow another player more space and increase the possibility of a goal. “… the large pitch, numerous players, limited player turnovers and sparse scoring” are some other issues in analyzing a match, as a result “soccer is arguably the most challenging to analyze of all the major team sports.” (Guiliang et al. 1531). Furthermore, data can be very expensive, making it difficult for the smaller clubs and leagues to afford (Decroos et al. 2). The setting is another important factor, and it changes in almost every situation, except for penalty kicks and corners. Therefore, it is no surprise that penalty kicks are the best analyzed part of soccer. One more limitation of data analytics is that most approaches cannot include the game context and often just include the player on the ball at a particular moment (Guiliang et al 1531).
Data analytics has four main advantages for the teams. To understand the following paragraph, one must be aware of the different kinds of data. Event stream data, which annotates the time and location of specific events and tracking data, records the locations of the players and the ball at a high frequency using optical tracking systems during the game (Decroos et al. 2).
The first big advantage is the data helping the team to recruit players. In contrast to the wealthy clubs, which spend millions of dollars for new players, the smaller clubs have more financial limitations, making them more dependent on a good recruitment process (FourFourTwo 3:57). A big help in the recruitment process is that there are no humans needed to visit every game of a potential player, but they collect data from different leagues and clubs, allowing them to get a pool of potential fits for their team. An analysis of the data shows the player’s performance, strengths and weaknesses (FourFourTwo 7:31). After they have the data from the various players, they look out for undervalued players, especially young ones with a lot of potential (FourFourTwo 6:24). Now the team has a few possible players, who are then watched by a scout. The data analysis is not replacing scouting, but it helps the team to gather more information about the players, having a bigger pool of potential fits and making the scouting process a lot faster. A good example to illustrate this process is Southampton, an English Premier League Team, which created a “black box”, a live database collecting player metrics from every major league (FourFourTwo 10:17). It enabled them to acquire players of undervalued talent and sell them on for a profit (FourFourTwo 10:25).
The second big advantage of data analytics is the evaluation of player’s performances in games and practices. The main goal is to get the maximum out of every single player (FourFourTwo 5:52). The data which helps to recruit players is also used in evaluating a player’s performances. Besides the performance of the individual player, analytic data shows the team’s performance as well. It allows the coach to have a better understanding of weaknesses and strengths, which allows him to improve the team’s weaknesses in practice. The data also helps him to decide which player to sub off or on during a game.
The third advantage of data analytics is the analysis of the opponent team. Even if the team has never faced their opponent or the opponent team has new players, data analytics allows a detailed analysis of the style of play, weaknesses and strengths of the opponent team and individual players. As the Assistant Manager of Forest Green Rovers FC Scott Lindsey said: “We may know more about the opposition than they actually know about themselves.” (FourFourTwo 13:01). The coach can prepare the team way better, with detailed instructions and a specific game plan. The pre-match analysis of the opposition became essential in the last few years. Nowadays, surprises in terms of the opposition’s tactics, match plan or starting formation are very rare. If the opposition’s team’s average position is quite deep, one conclusion is to possess the ball and reduce long balls. Today, players in the top leagues know the strengths and weaknesses of the player they are facing, allowing them to expect certain movements or runs.
But the use of artificial intelligence in terms of data analytics also helps the players to improve, which leads us to the last major advantage of data analytics. Injury prediction is the “probability of an injury from past and current season data”, which is one aspect that helps the player to improve (Tuyls et al. 16). Of course, the injury prediction helps a player because it allows him to adjust his workload and training, in order not to risk any injury. Especially international clubs have an intense game plan, with its many injured players can extremely reduce the team’s performance. The acute-chronic workload ratio serves as a predictor for sports related muscular injury (Tuyls et al. 16). Current injury prediction methods are limited by the availability of data, which makes the sports clubs take preventive measures as a result (Tuyls et al. 16). The prediction of injuries is a promising avenue for application for artificial intelligence techniques. As a player, continually improving your skills plays an immense role, which is another advantage of data analytics. Knowing weaknesses and strengths helps to build a specific training plan. Vision based player tracking, pose estimation and event detection can improve learning modeling, subsequently improving the player and theoretic analysis of game strategies (Tuyls et al. 8). In addition, the data also improves the quality and accuracy of feedback for players from computer vision techniques, which enables the reconstruction of real game scenarios (Tuyls et al. 7).
The development of artificial intelligence brings future models around soccer.
One model is the “deep Reinforcement Learning (DRL) model to learn an action-value Q-function” (Guiliang et al. 1531). The goal is to model the value of a certain action in a match, depending on the game situation, difficulty and outcome of the situation. The team’s chance of scoring the next goal depending on the game text is another model (Guiliang 153).
An interesting challenge is considering the current game score, the remaining time, the relative strength of two teams and the impact of the current game decisions on upcoming matches, play style, counterattack types and defense schemes (Tuyls et al. 15). Furthermore “soccer has not yet cracked the code in terms of what are the key indicators of what is going to make a player successful or not” (Brian Bliss, FourFourTwo 17:00). An Automated Assistant Coach could improve the understanding of human soccer and the overall experience of the game for players, coaches and spectators (Tuyls et al. 4). It could provide the coach with live data of the players and continuously updated in-game evaluation of the opponent team.
In conclusion, after some hesitation, Artificial Intelligence, especially in the form of data analytics, has found its way into the world of soccer. It helps teams with their recruitment, allows the coaches and staff to make a better analysis of their own team, players and their opponent teams and assists the players to maximize their potential. As a result, coaches can make their players better and they improve, making an overall better team. Therefore, it is very interesting, what the future of soccer will look like, and how the development of artificial intelligence is going to change the world of soccer.
Bibliography
Claudino, João G., et al. “Current Approaches to the use of Artificial Intelligence for Injury Risk Assessment and Performance Prediction in Team Sports: A Systematic Review.” Sports Medicine – Open, vol. 5, no. 1, 2019, pp. 1-12. ProQuest, https://wilkes.idm.oclc.org/login?url=https://www-proquest-com.wilkes.idm.oclc.org/scholarly-journals/current-approaches-use-artificial-intelligence/docview/2258055149/se-2, doi:http://dx.doi.org.wilkes.idm.oclc.org/10.1186/s40798-019-0202-3.
Decroos, Tom, et al. Actions Speak Louder than Goals: Valuing Player Actions in Soccer. Cornell University Library, arXiv.org, Ithaca, 2019. ProQuest, https://wilkes.idm.oclc.org/login?url=https://www-proquest-com.wilkes.idm.oclc.org/working-papers/actions-speak-louder-than-goals-valuing-player/docview/2071619597/se-2, doi:http://dx.doi.org.wilkes.idm.oclc.org/10.1145/3292500.3330758.
Guiliang, Liu, et al. “Deep Soccer Analytics: Learning an Action-Value Function for Evaluating Soccer Players.” Data Mining and Knowledge Discovery, vol. 34, no. 5, 2020, pp. 1531-1559. ProQuest, https://wilkes.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/deep-soccer-analytics-learning-action-value/docview/2441910672/se-2?accountid=62703, doi:http://dx.doi.org/10.1007/s10618-020-00705-9.
Luzum , Nathan, and Michael Model . “History and Background.” The Soccer Analytics Revolution, Duke University , 4 Dec. 2022, https://sites.duke.edu/socceranalyticsrevolution/history-and-background/#:~:text=Soccer%20was%20more%20resistant%20to,of%20players%20on%20the%20pitch.https://sites.duke.edu/socceranalyticsrevolution/history-and-background/#:~:text=Soccer%20was%20more%20resistant%20to,of%20players%20on%20the%20pitch.
“The Numbers Game. How Data is Changing Football. Documentary.” YouTube, uploaded by FourFourTwo, Dec 22, 2017, https://www.youtube.com/watch?v=lLcXH_4rwr4
Tuyls, Karl, et al. “Game Plan: What AI Can Do for Football, and What Football Can Do for AI.” Journal of Artificial Intelligence Research, vol. 71, 6 May 2021, pp. 41–88, 10.1613/jair.1.12505. Accessed 12 July 2021. https://dl.acm.org/doi/pdf/10.1613/jair.1.12505fff