Remember that soccer game where your favorite club came back from behind to win the match? Instead of spending time speculating about the match’s make-or-break moments—imagine knowing what plays led to victory, which players made the biggest impact, and how that yellow card from the first half affected the outcome.
getty
Machine learning (ML) is making this level of insight on matches possible for hundreds of millions of soccer fans globally. Organizations are using ML to understand, apply, and present their data in revolutionary ways to invent new experiences.
The Bundesliga—Germany’s premier national soccer league governed by the Deutsche Fußball Liga (DFL)—is paving the way for ML-powered innovation. The Bundesliga has transformed the game-day experience by using artificial intelligence (AI), ML, analytics, compute, database, and storage services on the cloud to generate in-depth, real-time strategic insights on soccer games—and bring remote fans closer to the action.
With machine learning, innovation is the name of the game
The Bundesliga regularly boasts the best average match day attendances in Europe. But when the global pandemic interrupted the league’s championship, requiring games to be played without an audience, the DFL faced a key challenge: better engaging with fans through screens by reinventing the remote fan experience.
With more than 500 million fans around the world, the Bundesliga is no stranger to engaging audiences across broadcast and digital channels. The league knew its fan base had an appetite for richer content that would bring them closer to the pitch. Considering soccer’s 90-minute matches are action-packed, the Bundesliga didn’t have to look elsewhere to provide this content. It just had to dig deeper into the game.
A single match produces approximately 3.6 million unique events, with each event having the potential to generate interesting insight. The ability to analyze these data points and relay insights can enrich storytelling in soccer, helping fans better understand how strategy, skill, and luck impact the game.
“Data can help create a much better fan experience for spectators in front of a television screen or iPad because it helps them engage with the game on a deeper level,” says Simon Rolfes, sporting director of the Bundesliga club Bayer 04 Leverkusen. “Fans want more information about the performance of their favorite players and teams, like how fast they are, what tactics they’re using, and the quality of playing.”
Achieving this level of insight would have been too cost-prohibitive five years ago and likely impossible 10 years ago. But advances AWS has made in deep learning over the past several years helped the Bundesliga make real-time match analysis a reality. Deep learning, a subset of ML, imitates the way our brains learn by processing data with artificial “neural networks” that can extract complicated relationships with little human supervision.
Reinventing the remote fan experience with AI, ML, and analytics
By partnering with AWS on their data strategy, and using analytics, ML, and other cloud services, the Bundesliga is providing a remote fan experience like no other, offering real-time data-driven insights about team and player performance in every game. These insights, called Bundesliga Match Facts, are the first of their kind: a unique combination of advanced statistics and game analyses that provides new insights into the action on the field.
“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 a better understanding of the split-second decisions made on the pitch,” says Andreas Heyden, CEO of DFL Digital Sports and EVP of Digital Innovation for the DFL Group. “The use of Bundesliga Match Facts enables viewers to gain a deeper insight into the key decisions in each match.”
To achieve Bundesliga Match Facts, each Bundesliga stadium is equipped with up to 20 position-tracking cameras. Deep learning-powered computer vision tracks player and ball movement and other events at a 25 Hz frame rate, and translates them into position data, event data, and metadata. The data is processed by advanced ML models to produce unique categories of Bundesliga Match Facts. Each ML model is trained on AWS Sagemaker by analyzing thousands of data points from previous seasons.
Using the cloud, Bundesliga Match Facts are instantly aggregated and distributed to broadcasters as well as Bundesliga’s platforms and channels. From start to finish, each Match Fact is calculated and distributed within 500 milliseconds—about 20-40 times faster than the time it takes for live video footage to reach the screen.
The data provides a level of understanding commentators and fans could previously only speculate about, like the probability of making an attempted shot, the pass strength of a given team, and even which players are pressured the most often.
“Data adds a different layer of storytelling,” says Heyden. “For example, maybe the home club is up 5-0 and scored the sixth goal in the 90th minute. It’s not a deciding goal, but if the commentator could say it was the most improbable goal this season because it had only a 2% chance of getting into the net, it can help enrich a fan’s appreciation of the game.”
ML also helps the Bundesliga captivate its audience beyond gameday through automated content production. With over 70 broadcasting licenses across 200 counties, the Bundesliga uses ML to generate audience-specific highlight reels. “The vast amount of our customer demands and fan insights would not be satisfied without the power of machine learning and the cloud,” says Heyden.
For example, the Bundesliga has a significant fanbase in Latin America that follows superstar players in the league. After the final game-day whistle, ML technology will create a video compilation featuring match highlights from these big-name players in seconds. This reel is then sent through the cloud and distributed to audiences across Latin America.
For the 2021-2022 season, Bundesliga has added another innovative service to its lineup. The Data Story Finder, developed on AWS using smart algorithms, accelerates the delivery of context-related live information to broadcast commentators. It correlates live match data captured automatically in real time with other match, seasonal or historical data, then presents the results to Bundesliga data editors as additional contextual information. Commentators can then share this supplemental data – such as surprising, unusual or new facts/accomplishments – to enrich the viewing experience. The DFL is the first in the world to offer an AI-supported live-commentary tool.
Sports fans will be able to see the DFL’s new technologies and innovation in action at SportsInnovation 2022, an annual trade show that showcases technologies from across global sports.
Key takeaways for business leaders generating predictive insights with ML models
The Bundesliga’s experience provides several best practices for other business leaders interested in using ML to enhance innovation:
- Embrace cloud-first strategies. Before tackling ML, the Bundesliga had to upgrade the legacy systems it used to store, process, and extract data. “Moving to the AWS cloud and freeing ourselves from the limitations of legacy systems was the first step to making real-time match analysis a reality. Now, every piece of content stored in our media & data hubs is instantly available dating back to 1963,” according to Heyden. With cloud computing, the Bundesliga can maximize data storage, accessibility, and efficiency.
- Scale as needed. The resources needed to deliver Bundesliga Match Facts are robust and only needed on matchdays, which is why Bundesliga uses scalable cloud services. This flexibility enables the Bundesliga to innovate its broadcast product when needed—and they can do this for a fraction of the cost of maintaining their own infrastructure.
- Work backwards from the customer. By understanding the interests of various stakeholders—fans, broadcasters, the press—the Bundesliga was able to deliver content that would most captivate their audience. This is a crucial lesson for business leaders: Start with the end-customer in mind and work backwards to create a solution that meets their needs. “The technical creation of a Bundesliga Match Fact is complicated, but it’s easy when you work on such elaborate technology,” says Heyden “The real challenge is thinking of the naming, the on-air design, the story to be told and making it relevant for the end customer. Working backwards from what the end customer really wants helped us achieve that goal.”
- Solicit feedback from the end user. The Bundesliga is always trying to improve the fan experience. By listening to audience feedback on Bundesliga Match Facts’ naming, presentation, and more, the league was able to adjust accordingly and improve the clarity of its content. “Fan feedback is critical and has helped us adjust several Bundesliga metrics to make them clearer and more accessible,” according to Heyden. These insights are also helping the Bundesliga determine what new capabilities to develop in the future.
- Prioritize alignment across the organization. The Bundesliga owes Bundesliga Match Facts’ success, in part, to alignment between its technology teams, business units, and key executive stakeholders. The biggest mistake companies make when undertaking technology and data initiatives is diving headfirst into technology without setting objectives and key results (OKRs). By sharing business OKRs with technology teams, you can deliver tech outcomes that drive business outcomes.
By leveraging existing data and embracing machine learning in innovative ways, the Bundesliga was able to reinvent the remote fan experience and find new revenue sources—and they’re just getting started. ML insights are also helping the Bundesliga clubs improve their preparation before games, determine which players to recruit, and provide more targeted training for its players.
Learn more about how other leading organizations are reinventing their business and redefining their industries with AWS.
