AWS for M&E Blog
Bundesliga Data Story Finder: Delivering fans the stories they love
Football is about passion and broadcasters use storytelling to transmit their football passion to the fans. Every moment, or insight, of a match could turn into an iconic story. Being able to extract meaningful insights from football’s ever-growing data landscape continues to intensify.
Capturing event data for significant moments in the game (such as shots, passes, tackles, and fouls), along with positional tracking, advanced stats, and ball data is now standard across professional leagues. With this amount of data, editorial teams and commentators face an overwhelming task: identifying the most relevant stories, during live matches while managing pre- and post-match analysis.
Bundesliga needed help sifting through mounds of data so they could effectively transmit their passion, and story making moments, to fans worldwide. Bundesliga captures more than 3.6 million data points during each match. They partnered with Amazon Web Services (AWS) to leverage this mountain of data as a foundation—unveiling edge-of-your-seat stories through Data Story Finder.
AWS and Bundesliga introduced Data Story Finder for the 2021/2022 season to address this challenge by automatically detecting compelling narratives from live match data. Data Story Finder enables Bundesliga to navigate through the data points in real-time using generative AI, and to correlate the data into football stories.
The machine learning capabilities of AWS generate Bundesliga Match Facts based on the 3.6 million data points captured each match. Data Story Finder evaluates the Bundesliga Match Facts against pre-defined story templates. Today’s insights (like comparing shot efficiencies and pressure analysis) are then put into the hands of commentators in real-time through the Commentary Live System.
The success of Data Story Finder has paved the way for providing more automated stories to partners, even beyond Bundesliga Match Facts data. Editorial teams at media partners, and various leagues, required deeper historical context and comparative insights that could further enrich the football storytelling capabilities. Data Story Finder has reduced the effort needed for analyzing the inevitably growing amount of match data.
Beyond Bundesliga Match Facts: Historical performance analysis
The second generation of Data Story Finder (version 2.0) expands its analysis scope beyond real-time match statistics—encompassing comprehensive historical performance data. This evolution enables the generation of stories, while putting them in the broader seasonal and career contexts—providing commentators with a richer narrative framework.
Figure 1: Commentary Live System interface displaying real-time story generation.
Three new story categories define this expanded capability:
- Event-based stories trigger automatically when significant match events occur, generating contextual information that spans the entire season. When a player scores, the system immediately calculates their position in the season’s top goal scorer rankings. A goalkeeper’s clean sheet streak is automatically tracked and reported when it ends, complete with details about the last goal conceded. Team performance patterns, such as conversion rates when taking early leads, are surfaced instantly.
- Exceptional performance detection identifies when players or teams exceed predetermined thresholds during matches. The system evaluates whether these performances represent personal bests, team records, or competition-wide achievements. A midfielder’s passing accuracy might trigger a story if it represents their season best, while a team’s possession percentage could generate content if it ranks among the competition’s top performances.
- Match summaries provide automated analysis at halftime and during key match moments, contextualizing the current game within historical patterns. These summaries, generated with the help of generative AI services on AWS, compare different time periods within matches and evaluate team performance against seasonal averages.
The distinctive capability of Data Story Finder 2.0 lies in automatically recognizing when live events become historically significant. Consider a player breaking the season speed record mid-match—Data Story Finder 2.0 instantly identifies this achievement and delivers the context through the Commentary Live System. This allows commentators to highlight the milestone as it happens.
Unlike traditional systems that provide only current match statistics, Data Story Finder 2.0 adds context from entire seasons and previous campaigns. It helps commentators, and media partners, tell richer stories without manual data analysis.
Technical architecture: Scalable and independent
AWS enables Data Story Finder 2.0 by leveraging core components from the Bundesliga Match Facts architecture, while introducing new scalability features. The system uses Amazon Managed Streaming for Kafka (Amazon MSK) to handle data ingestion, while a serverless architecture leverages Amazon EventBridge, Amazon DynamoDB and AWS Lambda to process each story type as separate workflows. This allows different stories to be processed independently, each having its own scalable resources and failure isolations.
Figure 2: High-level architecture diagram showing Data Story Finder 2.0 integration.
This architecture provides reliability through isolated processing workflows, meaning if one story type encounters an issue (like statistical calculation) it won’t disrupt the generation of other stories during live matches. Each story type scales automatically based on demand, facilitating consistent performance during peak periods of simultaneous game activity—like when several goals are scored across different games in the same minute.
The Data Story Finder 2.0 incorporates a lightweight framework that streamlines the integration of new story types, reducing development time and maintenance overhead. A common data layer maintains real-time updates about individual players and team performances at both match and season levels. This enables story generators to access current statistics in a standardized way without complex data queries.
The story generation logic remains modular, with each story type defining its own business rules and guardrails. This separation allows for independent updates and customization based on specific editorial requirements or league preferences.
Expanding distribution channels
Integration capabilities were extended beyond the Commentary Live System, with fan-facing live tickers included. Stories generated by Data Story Finder can be ingested directly into public-facing platforms (including official league websites).
For editorial teams using the Commentary Live System, the Story Monitor interface provides control over story publication, allowing editors to review, modify, or remove automatically generated content before it reaches commentators. This human oversight verifies editorial standards, while maintaining the efficiency benefits of automated story generation.
Fan-facing implementations operate with varying levels of editorial involvement. Some platforms integrate stories directly into their live feeds, while others provide editorial teams with story suggestions that can be adapted for their specific audiences and editorial guidelines.
The multilingual capabilities inherited from the original system support international distribution, with stories available in German and English. This serves diverse editorial teams and audiences across different markets.
Future development: Predictive analysis
Bundesliga, along with AWS, is constantly looking ahead, so Data Story Finder will continue to evolve—incorporating predictive capabilities inspired by how commentators prepare for matches. The system will analyze upcoming fixtures to identify potential storylines and prepare for likely significant moments before they occur. By evaluating team form, player trends, and historical matchups, Data Story Finder will anticipate records that might be broken and milestones that could be achieved.
The combination of reactive story detection with predictive analysis creates new possibilities in sports data utilization. By combining comprehensive historical analysis with real-time event processing, the Bundesliga Data Story Finder serves as a powerful companion to journalists and editors. It empowers them to continue delivering rich and engaging experiences to football fans, while expanding the depth and relevance of automatically generated sports narratives.
Contact an AWS Representative to know how we can help accelerate your business.