AI in sports: How modern tools are changing analytics
With recent technological advances making it easier to access and operationalize data, advanced analysis is now the norm for teams when evaluating players and plotting strategy.
Editor's note: This is the second article in a three-part series examining the evolution of AI in sports. Read Part 1 here.
After the use of analytics in sports spread over the first two decades of the 21st century, AI began to emerge as a means of analyzing data late in the 2010s.
Teams such as the Minnesota Twins in baseball and professional sports organizations such as the Women's Tennis Association invested in machine learning to do predictive modeling. But not all teams and leagues made the same commitment. Until the past few years, AI development took significant investments in technology infrastructure and personnel with the proper expertise that not everyone could make.
Now, however, just as generative AI (GenAI) reduced some of the barriers to AI development and led to rapidly increasing use of AI as a means of analyzing data in the enterprise, it is helping professional teams in all major sports expand what they can do with data to make personnel decisions, scout opponents and determine strategy.
"They're all doing it now," said Peter Zaimes, a senior lecturer at the University of New Hampshire and founder of the university's Sport Analytics Lab.
Although sports analytics has evolved and every team in each of the major sports leagues is using AI to inform decisions, it's not visible to the naked eye. Instead, it's in what the eye now sees when observing games.
The Oakland A's pioneering use of analytics in the early 2000s took place behind the scenes. Fans didn't see what went into decisions regarding trades and free-agent acquisitions, instead understanding that the team's roster would have been vastly different had analytics not been behind decisions.
As analytics in baseball and other sports became more sophisticated, it was behind decisions informing defensive shifts, passing in football and 3-pointers in basketball, and the evolution in hockey away from toughness in favor of speed.
AI is similarly in the background. AI in sports is about advancing analytics beyond its previous limitations. It's about access to more data by more people in more sophisticated ways to make better decisions.
It's in the scenario planning that has led to the recent surge in football coaches going for a first down on fourth down, rather than punting to the other team. It's in the increasing emphasis on load management in basketball, with players resting more -- even sitting out entire games -- so they're fresh for the postseason. It's in decisions related to pitch sequencing and lineup construction in baseball.
GenAI a game-changer
Perhaps the primary reason AI is now the main driver of analytics in sports is that advances in AI -- GenAI, in particular -- have made AI models and applications easier to build, according to Zaimes.
Before ChatGPT's 2022 launch and subsequent improvements by OpenAI and competing vendors such as Anthropic and Mistral AI, developing an AI model or application required backgrounds in computer science and data science, including extensive coding skills. Now, because of GenAI, people can build models and applications using natural language rather than code, greatly simplifying the process.
"The casual analytics person within front offices, who is an amateur coder at worst and learning how to do it better, [can build AI tools]," Zaimes said. "I've read a lot of research … about people building models who aren't coders, and they look like the computer science coding from two years ago. That's permeating through sports, which is going to be a game-changer."
Meanwhile, taking advantage of surging interest in AI development, many technology vendors have built tools and created ecosystems that simplify developing AI models and applications. For example, tech giants such as AWS and Microsoft, data platform vendors including Databricks and Snowflake, and even more specialized data management and analytics companies such as Domo and Qlik now provide environments that simplify AI development.
Numerous teams -- baseball's Minnesota Twins and Texas Rangers, the WNBA's Indiana Fever and soccer's Philadelphia Union -- use Databricks as their platform for data management and AI development. A platform like this enables teams to do predictive analysis, something they couldn't do before building AI models, according to Alexander Booth, a solutions architect at Databricks and formerly the Rangers' assistant director of R&D.
AI allows us to move from describing the past to predicting the future. Teams can simulate matchups, spot fatigue or injury risk and give coaches a clearer picture of what might happen next. This gives decision-makers confidence to act quickly rather than waiting until after the game to analyze what went wrong.
Alexander BoothSolutions architect, Databricks
"AI allows us to move from describing the past to predicting the future," he said. "Teams can simulate matchups, spot fatigue or injury risk and give coaches a clearer picture of what might happen next. This gives decision-makers confidence to act quickly rather than waiting until after the game to analyze what went wrong."
The technology that teams use to develop AI tools is the same as the technology enterprises use to develop applications that inform business decisions, Booth continued. However, AI is applied differently in sports, given the need to constantly make a multitude of minute decisions over the course of a game.
"The technology is the same, but in sports, decisions have to be faster and easier to explain to less technical stakeholders," Booth said. "Building trust with coaches and players is just as important as building the model itself. The pressure of game-time decisions forces teams to focus on clarity and usability in a way most businesses do not."
Applying AI
While technological advances have made AI tools easier to build, enabling all teams in major professional sports to advance their analytics capabilities beyond, no two teams use data the same way.
Each league has partnerships with companies that provide all teams with data and certain insights. For example, MLB uses Google Cloud Platform to deliver live game data and analysis under the Statcast name, while the NFL uses AWS to provide its teams with information and analysis under the Next Gen Stats umbrella.
However, what each team does with the information provided by their respective leagues is different, and that's where teams are using AI to try to gain a competitive edge.
The Union of Major League Soccer had no real data infrastructure when Addison Hunsicker, the team's senior manager of soccer analytics, was hired in 2020. In 2021, however, the team began using Databricks for data management, data modeling and application development.
MLS partners with Sportec Solutions to provide physical data such as the total distance run by each player during each game. The Union, meanwhile, gets event data such as who made a pass, to whom and how far the pass traveled from other providers such as SkillCorner.
Before adopting a data management platform, data from separate sources was isolated into different systems and was difficult to combine. Once integrated within Databricks, the Union was able to develop metrics that enabled the team to better understand players' off-the-ball movements.
"In soccer, so much of what a player does happens off the ball, so [it's important] to calculate player performance when they're not dribbling, passing or getting involved in a defensive action," Hunsicker said.
Using such information, the Union -- currently in first place in the Eastern Conference after finishing 12th in 2024 -- has been able to target players that fit its transition-oriented style of play, fueled by playing long passes to create a roster that fits well together.
Still, the Union's use of data was limited before AI. Now, the team is doing more predictive analytics than in the past thanks to AI, according to Hunsicker.
"That's how I use AI and implement it into my workflows," he said. "The big way that AI can play a role is by tying [predictive analytics] to the new metrics that are being developed. … We know that other teams in MLS are also partnered with SkillCorner and have access to the same information we do. Each team is going to use it differently, and that's where you can get your edge."
In addition, AI can help the Union
After analytics use in sports grew over the course of about two decades, recent advances in AI technology have enabled all teams in major sports to more easily make data part of the decision-making process.
better respond to questions from coaches and players, Hunsicker continued. For example, the Union built a chatbot using AI development tools from Databricks that enables users to ask questions about MLS' rules and regulations, including complex salary cap rules that influence the team's roster, and get immediate responses.
Meanwhile, other queries that might have required days of searching for relevant data, data modeling and ultimately analysis can now sometimes be answered in minutes.
"When I know what I want to ask, and I know the data I want to work with, and I know the answer is in there somewhere, it doesn't take too long," Hunsicker said. "If I need to ask more questions and figure out where I want it to head, then it might take a little bit longer."
Access to analytics
AI might be influencing all the major professional sports, but just as some teams used analytics more than others, not everyone is bought in on AI as a primary source for decisions. In addition, not everyone has access to all the data they'd like.
Players, in particular, don't always have access to the data they might want to analyze their own performance and potentially improve. They're dependent on the teams they play for to provide them with data, and collective bargaining agreements between player unions and leagues define what they can access.
In the NFL, players are provided with data about the opposition to help them prepare for games, according to Kelvin Beachum, an offensive tackle with the Arizona Cardinals. Beachum is beginning his 14th season in the NFL.
"As it pertains to evaluation and coaching and scouting, guys are using everything under the sun to gain an edge to be able to win every rep on Sundays," he said.
Players are not, however, provided with all the data related to their own performance, Beachum continued. For example, teams collect performance data using vendors such as Catapult, Zebra Technologies and Next Gen Stats from AWS. While they might share that data with a player on their team, they are not required to do so, and free agents do not have access to such data collected by previous employers.
"There is a baseline in terms of what teams provide you as you're preparing for the next game," Beachum said. "But today I still … don't have access to Zebra Technologies, Catapult technology, Next Gen Stats. Those are the three that are always collected on me, whether I want them to be or not. I can't get that information."
Like anyone, Beachum can collect data on his own to glean insights about health and performance, using technologies such as Whoop, Apple watches and Aura rings. But even if he did collect such data, it would be captured within each application with no way to integrate and assess it.
Other sports are more progressive in terms of what teams share with players, according to Beachum. The NBA is particularly advanced in terms of analyzing how the body reacts and performs over the course of a season and sharing that data, he said. The MLB is also more forward-thinking than the NFL when it comes to providing players with more than just data about the opposition.
Career length is one reason NFL players haven't collectively bargained to gain access to all the data captured on them. With the average career lasting fewer than four years, player unions tend to focus on other negotiating priorities.
"We're a very barbaric game; to an extent, we're somewhat archaic in our league," Beachum said. "It's not that we're slow, but we just haven't been able to adapt as fast as other sports have, in some respects. In the NFL, some guys don't ever get a look at a second contract, so what is it even worth making that type of investment [in gaining access to all data]. It's just different dynamics."
Acceptance
While some athletes, such as Beachum, want more data, others wish AI and analytics played a smaller role in sports, at least when it comes to how teams make strategic decisions and pass them on to their players.
The NFL, even as AI enables new types of analysis, has perhaps been the slowest major professional sports league to make analytics a pervasive part of on-field decision-making. Baseball, meanwhile, remains the most driven by analytics, in large part because so much of its action is centered around the one-on-one matchup between pitchers and batters.
Among baseball teams, the Dodgers, Yankees, Red Sox, Rays and Astros are known to be among the most reliant on analytics, with their use of analytics only growing as AI makes more data available and enables new types of analysis.
Not all players, however, care for their teams' increasing use of analytics.
Mets first baseman Pete Alonso, one of the top power hitters in the sport, in 2022 critiqued the devaluation of runs batted in, saying that baseball is a competition to score runs, not to get hits.
Former Yankees, including pitcher Zack Britton and outfielder Clint Frazier, have claimed that the team's increasing reliance on analytics has led to a disconnect between decision-making and the fundamental competitive nature of the game.
Max Scherzer, a three-time Cy Young Award winner and likely a future Hall of Famer, while a devotee of analytics to glean insights about pitching mechanics and the shape of pitches themselves, has also expressed distaste for data being the primary determinant of strategic decisions.
To foster buy-in, front offices in sports have to take the same approach to change management that enterprises do when expanding their use of analytics and AI, according to Databricks' Booth.
"The biggest challenge is making AI insights clear and reliable enough that coaches and players actually use them," he said "Teams also need to foster a culture of curiosity, where staff feel comfortable asking questions and experimenting with the data. Without that culture, even the best technology will sit unused on the sidelines."
At the crux of the disconnect between a team's embrace of analytics and a player's resistance is the perception that metrics are more meaningful than a player's competitive drive and ability to perform counter to what data might conclude, according to UNH's Zaimes.
If a batter has a poor history against a certain pitcher, or an offensive tackle such as Beachum struggles to stop a particular opposing defensive end each time their teams play, a team's inclination is to try using someone else. A player's inclination, however, is to never give up and continue trying to win the matchup.
"This generation of athletes is growing up with analytics, the professional athletes, but they didn't grow up with it," Zaimes said. "They're still like, 'I'm going to get the better of this guy, even though he got the better of me last time.' But there might be a front-office person who says, 'Hold on, we're going to sit you this game because of this matchup,' and then the athlete throws their hands up."
Whether all athletes embrace analytics or not, as AI continues to evolve and becomes easier to use, data-driven decision-making in sports is only going to grow in the coming years, according to Zaimes. And it will be the teams that best figure out how to use AI -- to incorporate analytics into decisions in ways players trust and accept -- that will find the greatest success.
"The winners and losers will depend on how they use it," Zaimes said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.
