AI in sports: how artificial intelligence is transforming the future of athletics

Contents:

1. How AI is Used in Sports Today
2. Analyzing Player Performance and Team Tactics
3. Improving Decision-Making with Real-Time Data
4. Predicting Outcomes and Performance Trends
5. Artificial Intelligence in Sports Training and Development
6. AI in Injury Prevention
7. How Professional Teams Use Advanced Analytics and AI in Sports
8. Innovation in Sports: Beyond the Playing Field
9. Enhancing the Fan Experience
10. Smart Stadiums and Connected Venues
11. Automated Content Creation and Highlights
12. AI in Popular Sports Around the World
13. Football (Soccer)
14. Basketball
15. American Football
16. Tennis
17. How AI is Changing Sports Betting and Predictions
18. Benefits and Challenges of AI Adoption in Sports
19. The Balance Between Human Expertise and Automation
20. Frequently Asked Questions
21. How is AI used in sports?
22. What is artificial intelligence in sports?
23. Can AI predict sports results?
24. Which sports use AI the most?
25. How does AI help prevent injuries?
26. Will AI replace coaches and scouts?

Artificial Intelligence has become a mainstay of industry around the world for decades, and sports are no exception. The most obvious use of AI in sports is to help athletes and their coaches perform in their very best form by processing a massive amount of information that would otherwise be impossible for a human to handle. This information could range from in-depth analysis of past performances and games, all the way to sophisticated predictive models that will provide athletes and coaches with the most up-to-date information. This allows them to make the best possible decisions in relation to their own performance and training, as well as that of their teammates, prior to a match or competition. In this guide, we will cover how AI systems use neural networks, as well as sophisticated algorithms, to allow for the best possible use in a wide array of sports. We will explore:

• How the information is collected
• How the information is then used
• Who uses the information, and in what ways
• The future of the human within the world of sport as information and technology continue to advance at a rapid pace

When looking at how AI is being used in sports today, it is important to recognize the difference between using technology to track information during games and actually gaining valuable insights into a team or athlete’s performance by using artificial intelligence. As more affordable solutions are hosted on cloud-based platforms, many different organisations are using a variety of AI systems, including youth training centres. This in-depth guide explores the use of a range of different neural networks and their algorithms, and looks at the applications for each within various sporting disciplines. It highlights the good and the bad, and also looks into the ethical side of using AI in sports before finally trying to achieve a balance between the automated analysis of information and the human element.

How AI is Used in Sports Today

The sports industry can be divided into many sub-industries. Currently, all of these sub-industries are going through a huge revolution with AI in sports. A huge amount of information is generated in each of the sports. This information can be processed by an integrated system that can generate the information required by an athlete or a team to gain a competitive edge. All the current analytics, computer vision, and tracking systems are being put to work in many different ways in the industry of sports.

Analyzing Player Performance and Team Tactics

AI is changing many of the ways in which sports are played. For example, a team of players can now be analysed in much greater detail than before. This is made possible by the high-resolution optical tracking of every player on the field as well as the ball. The positional data collected from this can then be stored for in-depth analysis of individual players as well as the team as a whole.

Many different forms of tactical behaviour can be automatically identified by feeding the data into a machine learning classifier. Behaviors that can be automatically identified include:

• How a team sets up when defending
• How they press the opposition when they don’t have possession
• How they deploy when trying to attack

All of these behaviours can be automatically detected by the AI, and the data from each behavior can be analysed in greater detail than a human analyst could ever hope to. For example, the AI can work out:

• The difficulty of a pass
• The amount of space between the defending team and the player with the ball
• The amount of space between teammates
• The speed at which a player is moving in transition

All of these factors can be given a numerical value to allow the AI to work out the “contextual” impact that a player is having on the game. This is much more than just the sum of an individual’s stats.

The data is then run through machine learning classifiers, for example, Deep Learning, that look at the data from previous matches and break down a team’s and player’s current and past tactics. For example:

• In football: the press
• In basketball on defence: the pick-and-roll
• In soccer: the team’s passing network

This allows for a multitude of factors to be taken into account when calculating a score for a player’s contribution in any given situation. The AI can give an exact score of a player’s contribution in any given situation, highlighting aspects of their game that would have gone amiss by typical statistics.

Improving Decision-Making with Real-Time Data

Historically, the coaches and trainers of sports teams have relied on a combination of past experience, observation, and even chance to make many of the key decisions that affect their athletes. However, thanks to the advances in the last decade of live data and advanced analytics, coaches today are able to make many of their key decisions with the aid of sophisticated software that processes the vast amounts of data being collected from a variety of different sources.

To cite an example: The AI application used by the Oakland Ballers, an independent baseball team, enables starting lineups as well as the substitution of relief pitchers during games. For now, however, advanced analytics and AI systems in sports are still strictly monitored by the various professional and amateur sports leagues. In these sports, the use of such technology is limited to certain types of data from various wearables, tracking systems, as well as other devices that provide biometric data. This data can then be displayed on the coach’s tablet on the sidelines of the game.

Predicting Outcomes and Performance Trends

Predictive Models for Future Events: When dealing with future events, and especially sports, there is a lot of data that can be used to train models to predict the most likely outcomes. In the case of football, for example, teams can use a number of different data sets—from team and player data to situations that have occurred in previous matches—to try and develop models that can forecast future results. The key here is to create a full map of all possible outcomes and make the best decision based on the results of the model, with the results representing the expected outcome. There are a number of different types of models that can be used in order to predict future events, some of which include development models of specific players.

On the other hand, models such as those developed for the 2022 FIFA World Cup are extremely effective for predicting outcomes on a population or season basis and attribute very little variance to individual games.

Artificial Intelligence in Sports Training and Development

The most important work that AI can do in sports is helping athletes train prior to a match. This can be achieved by using data from wearables that athletes currently use for training in new ways. These wearables currently include:

• Heart-rate monitors
• GPS
• Accelerometers
• Inertial measurement units (IMUs)

With a machine learning model, the volume and intensity of a player’s training, as well as the drills that they complete, can be automatically changed based on the physiological data of the athlete in previous training sessions.

The area of artificial intelligence in sport is changing the way that athletes are developed for sport. The way in which training sessions are delivered and how athletes train on an ongoing basis is experiencing a massive revolution. The old way of delivering training and how athletes trained was based on a coach’s experience and how they thought that the athlete should train in order to improve; however, the way in which athletes train for sport nowadays is vastly different. Wearable technology has grown massively over the last decade, and with this, there is a mass amount of data that is being created by the technology. The data that is being created is then being used by the athlete and their coach in order to create the best possible training for the athlete in order to improve their overall performance. Pose estimation is one of the key ways in which this technology is being used in order to provide the best possible training for the athlete, and is being used in a number of different sports and for a number of different types of training.

AI in Injury Prevention

There is a lot of hype around the ability of AI to predict with certainty when a player is going to get injured. However, currently, the best that has been achieved is to provide a highly accurate, probabilistic risk score for each player. This is created by aggregating a large number of different factors that have been shown to increase the risk of injury. These include a player’s current workload, as well as their previous medical history and a large number of other factors that have been shown to affect a player’s risk of injury.

Injury prediction systems have recently entered the market. The systems enable a team of coaches and physios to forecast the risk of injury for every player and for the team as a whole. The majority of systems that have recently entered the market utilise a player’s and team’s past injury data and use real-time data from every match to search for patterns that occur in the large data set.

Zone7 is an injury prediction system that was recently purchased by training load platform SportsCode. Zone7 aggregates large amounts of historical injury data from around the world along with a player’s and team’s past medical history, past workloads, as well as current workloads from every match in order to search for patterns and alerts—such as a sudden increase in high-intensity decelerations in training that have historically led to soft-tissue injuries in athletes. Getafe CF utilise Zone7 on a daily basis in real-time to allow their medical team to monitor all players on an ongoing basis in order to prevent overreaching a “red zone” where the risk of injury increases significantly. If a player were to reach a “red zone,” the coach would be alerted and make a decision on whether to take the player off training for a number of days in order for the athlete to allow their body to recover fully before reaching a point where serious injury could occur.

How Professional Teams Use Advanced Analytics and AI in Sports

In addition to the uses outlined for typical sports analytics, the way in which professional teams use advanced analytics and AI is different from typical analysis. Most types of sports analytics can be used for simple regression-based predictions based off of a team’s or player’s box score statistics, whereas advanced analytics and AI have become far more complex, as they use Deep Learning and other types of AI in order to create complex systems of similarity-based search and Natural Language Processing (NLP)-based automated evaluations.

The work of a team of data scientists within an elite sports franchise’s sports science department is to develop models that help the club’s management team simulate scenarios in order to provide them with insights on how best to utilise the salary cap. The models help them to establish an expected value for signing a free agent, and this is calculated by the expected number of wins the team will gain from that player over the course of a three-year contract. As an example of the sort of work that an integrated analytics department carries out, an automated video tagging system could be used to produce an opponent’s scouting report by automatically tagging a video of the opposing team using Computer Vision. The work of the data analysts would be greatly aided by such a system, as it would save them a lot of time by not having to manually clip together all of the relevant parts of a match.

Innovation in Sports: Beyond the Playing Field

The use of sports-related analytics and sports-related AI to improve the future of sports, beyond the work of individual athletes, is also transforming how individual teams operate their respective venues and deliver content to fans around the globe.

Enhancing the Fan Experience

Clubs can use data to increase engagement with fans in a variety of ways:

• Recommendation Systems: Can suggest personalised content, including videos of previous matches, highlights of favourite players, etc.
• Generative AI and Natural Language Processing (NLP): Can be used to create fan-facing chatbots and virtual assistants. This can handle tasks such as ticketing and answering other questions the fan may have.
• Augmented Reality (AR): Fans can be provided with an AR “stadium overlay” on their mobile using live data from the club. This would contain the same information that the TV presenter would be saying at any time and would allow fans who are not in the stadium to have a similar experience.

Smart Stadiums and Connected Venues

Smart stadiums and connected venues of the future are much more than just venues for sports. They are venues of experiences, business, data, services, and connectivity. Key features include:

• Complex models to forecast demand in order to optimise the pricing of tickets, food, and beverages.
• Computer Vision networks installed in venues in order to monitor and analyse large crowds of people.
• Facial recognition installed to enable fans to enter and pay for food and beverages without the need to physically interact with staff.
• Real-time monitoring to enable long lines to be monitored in real-time in order to identify and address problems before they become a major issue.

As with many other uses of AI, there are also many legal issues that need to be addressed, including the monitoring of fans’ personal information as well as the use of biometric data for surveillance.

Automated Content Creation and Highlights

Sports organisations and other media outlets are finding the work of analysing and distributing sports content much easier with technology. Computer Vision and audio recognition are key to automatically tagging highlights of games and even providing a detailed analysis of individual events. For example, an automated highlight package can be created within seconds of a series of key plays taking place. These plays could be big plays, scoring plays, or even plays that elicit a lot of crowd reaction, such as a loud cheer and a standing ovation after a tremendous slam dunk by a basketball player. The computer automatically tags these key moments of the game and creates a highlight package which can then be distributed immediately to fans through various media outlets.

Automated written recaps of games are also becoming more and more popular. These tools, known as Natural Language Generation (NLG), automatically read through live data feeds and then generate a pre-written recap of the action of the game. The recaps can then be distributed immediately to fans on the organisation’s website or social media outlets.

AI in Popular Sports Around the World

While the same machine learning can be applied to many sports, the effect that it has on different sports can vary due to the rules, the pace of play, and the environment in which the sport is played.

Football (Soccer)

AI in Soccer – Off and on the Pitch. From officiating to the analysis and planning of the match for coaches and their teams. Semi-automated offside technology uses computer vision to track 29 points on every player on the field. This information is then used by the Video Assistant Referees to make the most accurate possible decisions when reviewing incidents for potential errors. On the other hand, coaches of soccer teams can gain better insights into what their team is doing on the field. They can study the passing patterns of their team to analyse the space on the field that they control. By using tracking data of their players, the coach of a team can also gain deep insights into the load management of his individual players. This allows the coach to make the best possible decisions with regards to the rotation of his players on the field in order to prevent injury to individual players. Tools such as Zone7 can alert the coach of a team to possible areas of risk in order for the coach to plan and prepare his team for matches to come.

Basketball

Another way optical data is being put to use is in the creation of advanced shot quality models by Machine Learning (ML) that score shots taken on the court with great detail. Teams and players can then use this information to better manage their lineups and create the optimal spacing to create scoring opportunities. In addition, load management, similar to what teams do with their players’ minutes, can be done using AI like Zone7 to keep players fresh by ensuring that they are well rested prior to their next game. This is especially true during parts of the season where the team is playing a lot of games in a short amount of time. This can go a long way in preventing a variety of soft-tissue injuries, including muscle strains, that can affect players in the NBA due to the grueling schedule that they are required to play.

American Football

American football requires the most amount of data in terms of the amount of information required for the analysis of such data for sporting purposes. While all sports are constantly changing, due to the nature of the grueling sport, information must be available constantly to aid coaches and trainers to make decisions for their team in the best way possible. This information can be provided by the NFL’s Next Gen Stats.

Next Gen Stats is made up of RFID readers and Bluetooth Low Energy (BLE) beacons that are placed throughout stadiums. The majority of the RFID tags are placed in the players’ shoulder pads, while others are placed inside of footballs. These systems are able to provide a vast amount of biometric data as well as data with regards to where players are on the field at any given time. This information can then be analysed using machine learning and provided by the system for a number of purposes. Some of the things that can be done with this data include:

• Mapping out the routes of players
• Determining the degree of separation between a defender and an offensive player
• Determining the speed of a player
• Determining whether a team is running a pass play or a run play

On the defensive side of the ball, for example, a team could utilise a predictive algorithm in order to be able to determine the tendencies of play calling by the opposing team’s offense. College football programs are also using search algorithms in order to scour through the transfer portal in search of players who would be good additions to a program.

Tennis

Tennis is an individual sport. As such, it is not very open to error. That is because each player is solely responsible for their own physical skills as well as for their individual tennis strategies. Automatic content generation therefore focuses primarily on the tennis players’ biomechanics and on the spatial aspects of tennis. For this reason, Computer Vision is used to analyse all of the shots hit by both players during matches.

From the data gathered by this form of automatic content generation, it is possible to automatically calculate the trajectories of shots hit by tennis players as well as the amount of spin that they contain. In addition, it is also possible to determine the exact location on the tennis court where each shot landed. Automated content creation can also automatically predict where shots will land, and it can automatically highlight returns of the ball that have been hit by an opponent. The system uses the opponent’s footwork and their position on the tennis court to do this. This information can then be used by a tennis player or their coach in order to work on their tennis serving, as well as to provide advice on the stroke mechanics that the tennis player uses for each shot that they hit.

How AI is Changing Sports Betting and Predictions

As the world of wagering continues to go global, the use of AI in sports betting and predictions is increasing at an exponential rate. As the betting industry continues to expand to every corner of the globe, the use of AI in the commercial betting industry is being used in massive machine learning models that are created to attempt to predict the most likely outcome of a match. However, there is a misconception that the AI created by the sportsbooks can guarantee accurate predictions for the consumer. These models are calculated to create the most likely outcome of a match in order for the sportsbook to be able to manage their risk in order to guarantee a return on investment in the long run. This means that while the AI can accurately compute the win/loss probability for each match, the outcome of the match can fluctuate due to natural variation, refereeing inconsistency, injury to a key player, etc.

This information is not guaranteed to be 100% accurate and can actually result in a large amount of money for the operator, based on the model’s calculations using all of the data that has been input by the operator and more. The model processes this information constantly in order to produce win/loss percentages for future events, in order to update the lines and in-play odds for customers placing wagers on the operator’s site for future sporting events.

Benefits and Challenges of AI Adoption in Sports

While there are many benefits of AI in sports, there are also many challenges of AI adoption in sports. For the most part, these challenges can be put into two categories:

• The challenges of data privacy in sports and the use of information that is collected by tracking and athlete monitoring in sports.
• The many ways in which the technology can fail in order to provide negative outcomes for the athlete.

Just as there are benefits and challenges of AI in business, there are great benefits of using AI in sports as well as many challenges currently faced by the industry. One of the biggest current problems that is affecting the adoption of technology such as AI in sports is data privacy. There are many sports using a large amount of players’ and teams’ biometrics, and using them to increase performance as well as to provide fans with a greater experience of following the sports. The biggest issue with this is: who owns the rights to the data collected from the technology? As with many current issues with the use of technology, there is a lot of fear that the data could be exploited by external organisations. This is already a problem that has been seen with the use of facial recognition in recent years.

“Black box” AI modelling is another problem that is currently affecting many coaches who are using AI in sports. The “black box” refers to a situation in which a computer program is making decisions, but the human cannot see how the program has come to that decision. This can lead to coaches not understanding why the AI has made a particular choice, such as making a substitution. There is also a fear that there is a risk of biased AI modelling, which could cause problems for the recruitment of athletes of a certain type.

The Balance Between Human Expertise and Automation

There is a lot of hype surrounding the use of AI for prediction in many sports, and while there are certainly many applications of AI in sports, the fact of the matter is that most of them are utilised as decision support systems, meaning that the human is the end decision-maker. This is particularly true when it comes to matters that are inherently human, such as judging the merits of a prospective recruit based off of work ethic, leadership ability, and other intangibles that the algorithms cannot quantify.

As for how much of the decision-making process can be automated by the use of Machine Learning and AI, there are two schools of thought. The first is that the whole of a coach’s role could be taken over by a computer, and the other is that there are some elements which are too subjective to ever be taken over by a computer. The truth, as always, lies somewhere in the middle. Until the technology advances to the point where a computer can actually go and watch a youth match and assess the work ethic of a young player, their potential for improvement, their attitude to training in the off-season, their leadership skills, and their influence in the dressing room, it is likely that automated systems will always alert a coach to a player of high potential and then the human element takes over in terms of evaluation. The same would be true for a coach who uses such a system during a match to suggest a change in tactics, for example, an aggressive change of tactics in order to gain back a deficit on goal; the human element would then come into play and the coach would consider factors such as the current emotional state of the match, and the current momentum of the game.

Frequently Asked Questions

How is AI used in sports?

Currently in sports, AI can be used to improve the performance of athletes, aid the training of athletes, and help coaches during matches and the running of sports stadia using Machine Learning, Computer Vision, and Predictive Analytics to analyse large datasets of information and produce actionable information. This could be for the automatic generation of media, for real-time changes to teams during matches, or to inform and influence the training loads of athletes in order to improve their performance.

What is artificial intelligence in sports?

Artificial intelligence (AI) in sports refers to the use of advanced computer systems and machine learning algorithms to analyse data, identify patterns, and support decision-making across various areas of sport. By processing large volumes of information, AI helps teams, coaches, athletes, and organisations gain insights that would be difficult to obtain through traditional analysis alone.

Can AI predict sports results?

Using historical data, or time-series data for example, a system can predict with great accuracy win probabilities and trends in a given competition. As mentioned above however, there is always a lot of variance in sports, in addition to external factors such as weather and human error. This means that no system can guarantee a win in a given match, although the system can give an estimate of chances of winning.

Which sports use AI the most?

There are a number of sports currently using AI technology, but they are mainly spectator sports with a lot of money behind them. The “big four” sports of football (soccer), basketball, American football, and baseball are the sports currently using optical tracking, player and stadium sensor data, and statistics to help improve the performance of their athletes and teams. Other sports are using computer vision in training and teaching, such as tennis and golf, to provide the most in-depth and precise biomechanical analysis of a player’s movements.

How does AI help prevent injuries?

Workload from previous matches, biomechanical metrics from previous training sessions, and historical medical information of the athlete can be inputted into a machine learning survival model. The resulting risk scores, or predicted probability of injury for the athlete, can then be used by the athletes’ and teams’ medical staff to inform training sessions and ensure that the athlete is not putting themselves at risk of a soft-tissue injury prior to actual symptoms of fatigue appearing.

Will AI replace coaches and scouts?

No. Current technology is able to provide information that has been pre-filtered for the coach or scout. However, there will always be information that the computer has not been able to figure out. Therefore, the coach or scout will still be able to focus on the information that really matters, for example, a player’s emotional intelligence, work ethic, leadership ability, potential to create a positive team culture, and ability to form good working relationships with fellow players and coaching staff. In the end, it is up to the coach or scout to make a decision based on a variety of information, including the information provided by the technology.

See also: Top sports bettors of all time: famous professional gamblers