The premier German soccer league, the Bundesliga, recently announced increased collaboration with Amazon Web Services (AWS). What benefits is “The Beautiful Game” looking to gain from AI and machine learning (ML)?
Soccer is played and followed in over 200 countries around the world, making it the most popular sport in the world. It has evolved from games played primarily for entertainment and leisure to an industry rivaling any other in size and power. The English Premier League, the most popular soccer league in the world, has a total market value of approximately $10.7 billion. The Bundesliga comes in fourth at $5.4 billion.
To capture a slice of this growing market, frequent evolution is necessary to improve not only player performance but also fan enjoyment to get ahead of the competition.
One avenue for innovation lies in parsing vast amounts of data. This is where Amazon Web Services (AWS) steps in. “AWS is working with the world’s most renowned sports leagues to better understand their data and innovate upon it using our deep portfolio of machine learning services,” said Andy Jassy, CEO of AWS.
They are also working with the National Hockey League (NHL). (Read more about the AWS NHL story here.)
New Match Facts to Supplement Bundesliga Broadcasts
The Bundesliga became the first to deliver a combination of real-time advanced statistics and game analysis to soccer fans around the world. Bundesliga Match Facts (BMF) officially launched at the end of May 2020 with Expected Goals (the probability of a player scoring from a certain position) and Speed Alert (a display showing how fast a player can sprint). The league plans to leverage AWS’s cloud services, including ML and analytics, to more efficiently use its data, including introducing a new set of advanced statistics.
Of the three new BMFs, the Most Pressed Player is arguably the most innovative and interesting. It measures how often a particular player is put under significant pressure by the opposing team, and then provides a pressure rating for that player. The stat considers variables including the number of defending players as well as their distance and direction of movement relative to the player who possesses the ball. If a particular threshold is reached, the ML algorithm registers a pressure situation for the player in possession. It is then possible to determine which players on a team are put under pressure more often as compared to their teammates.
Attacking Zones visualizes the areas in which the teams push forward into the box. This involves dividing the final third of the pitch into four equally sized zones. Any time attacking players move the ball into one of these zones, whether through dribbling or making a pass, the algorithm registers an attack for the corresponding zone. On-screen graphics show how each team’s attempts to press forward are distributed across the four attacking zones as percentages.
Average Positions shows the average position of players on the pitch. It was introduced in the 2019–20 season. With the new Average Positions: Trends, this information can be evaluated for selected sections of a match and compared. This enables a visualization of how changes to a team’s tactical formation can impact a match’s outcome, such as scoring or conceding a goal, making a substitution, or losing a player to a red card.
Benefits of Match Facts to Players, Teams and Fans
The BMFs are calculable thanks to advances in technology that have enabled high quality tracking of positional data. Each one of the Bundesliga soccer stadiums is equipped with up to 20 cameras for automatic optical tracking of players, referees and ball positions throughout the entire match 25 times per second. As a result, the Deutsche Fußball Liga (DFL), which is responsible for running and marketing the Bundesliga, collects approximately 3.2 million to 3.6 million positions every match. These positional data points can then be used to construct a map of all players and the ball on the pitch at any given moment in time.
All match events, such as penalty kicks and shots on goals, are documented live and sent to the DFL systems for remote verification. Human annotators categorize and supplement events with additional situation-specific information. For example, they add player and team assignments and the type of shot taken (such as blocking or assisting).
One example of live player tracking and positional analysis can be taken from the match between FC Bayern (FCB) and SV Werder Bremen (SVW) on November 21, 2020. During the game, Leroy Sané, a winger for FCB, received the ball in the midfield under significant pressure, which is calculated over time based on the actions of Sané and opposition players.
While this definitely provides an interesting graphic for the fan, it is also beneficial for opposition teams. They can analyze this data and prepare for how best to deal with Sané when they face him during similar match situations. This is just one example of how BMFs such as pressure ratings can be used to better understand and evolve the game.
How Bundesliga Is Implementing the AWS Infrastructure
All these statistics are the result of a robust cloud infrastructure that delivers efficient and accurate data processing. The raw match data is fed into the BMF system on AWS to calculate the BMF values, which are then distributed worldwide for broadcasting.
As an example, to calculate xGoals (Expected Goals), an algorithm was created for the Amazon SageMaker to train an ML model on over 40,000 historical shots on goal in the Bundesliga since 2017. In addition to training the ML model using SageMaker, other infrastructure components and application-specific cloud components are integrated to handle the full cloud ML pipeline. This consists of:
- Data integration
- Data cleaning
- Data preprocessing
- ML model training and deployment
Data is ingested in two separate ways: AWS Fargate is used for receiving positional and event data streams, and Amazon API Gateway is used for receiving additional metadata such as team compositions and player names. This incoming data triggers an AWS Lambda function. The Lambda function takes care of a variety of short-lived, one-time tasks: automatic deprovisioning of idle resources, data preprocessing, and several data quality tests that occur every time new match data is consumed. Lambda is also used to invoke the Amazon SageMaker endpoint to retrieve the xGoals predictions.
Two databases are used to store the match stats: Amazon DynamoDB, a key-value database, and Amazon DocumentDB, a document database. For central storage of the official match data, Amazon Simple Storage Service (Amazon S3) is used. S3 stores the historical data from all matches and helps iteratively improve the xGoals model. It also stores metadata on model performance, model monitoring, and security metrics.
To monitor the performance of the application, an AWS Amplify web application is used. It gives the operations team an overview of the system health and status of Match Facts calculations and its underlying cloud infrastructure in the form of a user-friendly dashboard. This dashboard also helps collect metrics and continuously monitors relevant key performance indicators (KPIs), such as overall system load and performance, end-to-end latency, and other nonfunctional requirements.
The Goal of the AWS and Bundesliga Collaboration
AI and ML will not only be able to predict future plays and game outcomes, they will also recommend personalized match footage for fans across mobile, streaming and television broadcasts.
The Bundesliga is looking to increase engagement with fans and enhance the broadcast production. By using AWS Media Services, the games are not only being streamed live, but time-shifted as well within the event stream on the IP-enabled device that viewers prefer. Using the time-shift feature, viewers can pause and forward the stream, offering them more flexibility and interactivity.
On top of that, the league estimates that its partnership with AWS has resulted in operational costs reductions of 75 percent thanks to this highly scalable serverless streaming infrastructure.
“Working closely with AWS significantly enhances the investment we’ve made in innovation over the past two decades, all of which contributes to us being able to deliver a world-class football (soccer) experience for our fans,” explained Christian Seifert, CEO of Bundesliga.
“We at Bundesliga are able to use this advanced technology from AWS, including statistics, analytics and machine learning, to interpret the data and deliver more in-depth insight and better understanding of the split-second decisions made on the pitch,” added Andreas Heyden, executive vice president of Digital Innovations for the DFL Group. “The use of Bundesliga Match Facts enables viewers to gain a deeper insight into the key decisions in each match.”
The fans are no longer passive consumers of the sport. Teams and league management can make more informed decisions and cut down on operational costs while increasing fan engagement. However, these cutting-edge developments are only the beginning. It will be exciting to see what new innovations the AWS and the Bundesliga come up with to make the beautiful game an even more wonderful confluence of science and art. However, this is only the beginning. Sports fans around the world are looking forward to what new directions AWS will help sports take in the future. For a further look into that future, read this.