A seemingly simple technical gesture. Lots of composure. Thousands of reactions on social media. This is the type of event that LaLiga Tech analyzes when a player like Karim Benzema scores a penalty Panenka at 81.i minutes from the Champions League semi-final between Real Madrid and Manchester City on Tuesday 26 April 2022.
LaLiga Tech is an entity emanating from LaLiga, the organization responsible for the championship of the first Spanish football division. For almost eight years, it has been developing a set of technological solutions, including an OTT (over-the-top) platform that allows content to be distributed through a mobile or web application.
“In 2013, we made a big technological commitment to modernize and professionalize the competition and expand LaLiga’s growth opportunities,” said Tomm Woods, head of marketing and communications at LaLiga Tech. “It’s something that has driven LaLiga’s growth for many years around the world and in all aspects of its commitment: from how it produces its content, to how teams prepare, to the direction of the competition.” , add.
About 10 months ago, it launched LaLiga Tech as a technology subsidiary aimed at sports clubs that seek to attract fans, monitor player performance and protect the broadcasting rights of events of all kinds.
One of the specialties of LaLiga Tech is data processing. Whether it’s to analyze audience behavior, measure content consumption on broadcast platforms or social media, detect diversion of authorized streams and scams in sports betting, sell tickets and spin-offs, and more. “In the end, it all comes down to the data,” says Tomm Woods.
But there is no doubt that LaLiga Tech will create new silos or disperse through different systems. Therefore, data scientists have pushed for the adoption of a data lake.
“It’s easier to work when you have the data in one place,” says Rafa Zambrano, head of data science at LaLiga Tech. “When we started, it was complex to deploy statistical and machine learning models because the data was scattered across different systems.”
Rafa ZambranoData Science Manager, LaLiga Tech
LaLiga Tech chose Apache Spark, then Databricks
About four years ago, LaLiga chose to deploy a data lake. According to Rafa Zambrano, data scientists wanted “low-cost administration and maintenance” and sought to “accelerate the development” of their projects. Its choice fell to Databricks in the Microsoft Azure cloud through the Azure Databricks offering.
In a blog post (in Spanish), Guillermo Roldán, head of architecture at LaLiga Tech, explains in detail the choice of Databricks. The subsidiary wanted a batch or real-time event processing solution that would run in the cloud and be able to rely on GPUs. It was intended to avoid proprietary confinement, leaving the possibility of negotiating migration to other clouds if necessary.
Finally, the lead architect values the integration of Azure Databricks with Azure AD, scalable, pay-per-use environments, and data science environments integrated into laptops. Please note that Microsoft is one of LaLiga’s partners.
Despite the availability of Databricks in different clouds, vigilance is maintained regarding the dependency of the chosen vendor and publisher. To limit the effects, LaLiga Tech has opted to “bet on Apache Spark”, which gives it “peace of mind” as it is not necessary to use Databricks or Azure to run Spark’s work. , according to Guillermo Roldán.
That doesn’t stop LaLiga Tech from taking advantage of the vast majority of features that this “Lakehouse” offers.
“We use Databricks for a variety of big data projects, when we need the Apache Spark engine to process data quickly, we use the scheduler to automate and test our work, we have multiple metadata stores to manage data and security,” Rafa Zambrano. list. “We are processing real-time data with Databricks and are beginning to test MLOps and AI model governance features as part of our data science projects.”
Thus, the data lake gathers data from many DBMS to analyze the commitments and needs of more than 15 million fans on different websites, social networks, OTT platforms and in various applications.
“Having that data in a central data lake gives us a chance to really understand who the fans are and identify the profiles,” says Tomm Woods. “Is this someone coming to the stadium? Do you intend to buy derivatives? Who is interacting with our gaming platform ?, etc.”
More and more concrete data analysis
“It helps us reach the audience, to make the competition more relevant to this type of audience. And that helps drive the growth of our group, ”he continues.
The data science environment also includes data and statistics produced during football matches. “We have a vision of all the events of a match: results, fouls, number of passes, actions of each player, etc. as well as the monitoring of the position of the players and the ball by means of specific cameras ”, explains Rafa Zambrano. “We generate a data set of approximately 3 million lines per game.”
Each stadium of LaLiga member clubs is equipped with 16 optical tracking cameras that place athletes and referees in relation to the ball 25 times per second. To do this, LaLiga Tech uses computer vision technology from ChyronHego, a specialist in multimedia event broadcasting technologies. These data are combined with events obtained through the Opta Sport sports statistics platform.
This large volume of data obtained in batches or in real time allows “producing interesting statistics and predictions,” according to the head of data science. “For example, when a player shoots at goal, we are able to calculate the probability that he will score,” he illustrates.
This is one of the features of MediaCoach, a solution that is offered to both broadcasters and football clubs. This extracts more than 1900 data points per player (and 300 metrics in real time). Clubs use it to analyze and improve team tactics. The tool can also analyze the performance of the referees, in order to avoid interruptions as much as possible so that the matches are as smooth as possible.
To detect and prevent match-fixing, LaLiga Tech analyzes real-time data from forty sports betting platforms. A neural network is used to detect whether certain bets are out of the norm or to identify bettors close to teams or players. “Our neural network makes predictions about predictions. A regression model compares these predictions with real-time data from betting platforms,” explains Rafa Zambano. “The regression model makes it easier to explain the comparisons. We can indicate which variables most influence an output result before the team in charge of this analysis notifies the police.”
LaLiga Tech interested in other sports
Thus, LaLiga Tech has managed to develop a wide range of solutions to process the data of a sports competition: football. He now has to adapt his algorithms to other sports.
Rafa ZambranoData Science Manager, LaLiga Tech
“Right now my main challenge is to replicate what we have achieved with LaLiga to meet the needs of other competitions,” says Rafa Zambano. “We have to adapt our solutions to other sports, work with other partners.”
For his part, Tomm Woods believes that LaLiga Tech has been right in its technological commitment and that the entity, like its parent company, contributes to the attractiveness of the Spanish championship.
“We consider ourselves responsible for fan involvement, content protection, the fight against fraud and many other aspects,” says Tomm Woods.
“There are other institutions in the sports sector that want to do the same. We have technology packages that are already working, the right connectors for computer ecosystems and some clubs are using them in production, “he continues.
And if the first customers of LaLiga Tech are the Spanish and Belgian football leagues, the entity has every intention of offering its services to a set of sports leagues, including those dedicated to e-sports, but also to diffusers. “We have a contract with the Spanish Paddle League, a sport that is rapidly gaining popularity,” said Tomm Woods. For those who discover it, paddle tennis is a racquet sport that mixes the rules of tennis, squash, table tennis and Basque ball. “We built its OTT platform. It currently has more than 400,000 registered users,” says the communications manager.