Pragmatic advances such as improved data integration and streamlined information delivery will aid teams, as will transformative gains including agent-driven game strategies.
Published: 28 Aug 2025
This is the third article in a three-part series examining the evolution of analytics in sports. Read Part 1 and Part 2 here.
The slow growth of analytics over the course of a couple of decades, including some initial uses of AI, marked the past state of data-driven decision-making in sports.
Widespread use of AI to inform roster construction, player preparation and strategic planning marks the present use of analytics in sports. Generative AI (GenAI), in particular, is enabling access to more data by more people and making it easier to build machine learning models and other advanced applications that previously required highly specialized skills.
What happens next, and exactly how advances in AI -- such as the emergence of agents over the past year -- will transform the ways teams operationalize their data, remains to be seen.
There are pragmatic ways AI will likely help teams, such as better enabling them to integrate data collected from different sources and isolated in separate systems, as well as allowing them to create metrics and analyses that were previously too cumbersome. There are also most likely ways quickly advancing AI technology might transform the on-field product, according to Peter Zaimes, a senior lecturer at the University of New Hampshire and founder of UNH's Sports Analytics Lab.
"A lot of the creativity going forward is going to come from AI," he said.
For example, the offensive and defensive schemes football coaches come up with might be AI-generated, Zaimes continued. So might substitution patterns in basketball, forming forward lines and defensive pairs in hockey and pitch sequencing in baseball.
"It's all-around things that coaches haven't thought of before and are going to be told through AI," Zaimes said.
Practical progress
While AI conjures up science fiction tropes such as robots walking among humans and computers turning on their creators, the reality -- at least in the present and foreseeable future -- is far less dystopian. Instead, AI is being used to make people better informed and more efficient.
In the business world, enterprises are building chatbots and agents aimed at automating repetitive tasks such as documentation and summarization, as well as enabling more than just trained experts to use data to inform decisions and actions.
In sports, pragmatic ideas about how to use AI to advance analytics are similarly prevalent among teams and players. While there are fantasies about AI's potential to profoundly alter analytics in sports, with widespread use of AI still nascent, making information more available to improve decision-making is a main focus, according to Kelvin Beachum, an offensive tackle with the Arizona Cardinals now starting his 14th season in the NFL.
Analytics was virtually nonexistent in the NFL when Beachum broke in with the Pittsburgh Steelers in 2012. Late in the 2010s, teams started using data to scout opponents and optimize training methods to better prepare players for the grueling nature of a football season. More recently, analytics is being used to inform situational strategy, such as when to punt on fourth down and when to try for a first down; attempts on fourth-and-3 or shorter have steadily risen since 2017.
However, much of the data that NFL teams and players use is isolated in separate systems.
Teams are provided some data from the NFL itself, including Next Gen Stats from AWS. In addition, they collect performance data on players using wearable technology from vendors including Zebra Technologies and Catapult, neurological data to treat and prevent brain injuries and information related to players' recoveries after strenuous workouts and games.
Much of the data is unstructured -- text, images, videos, etc. -- which makes it difficult to organize and transform so it can be combined with other data. As a result, teams are often unable to get complete views of their players without toggling between separate systems, and players never get a full understanding of how their bodies are reacting over the course of seasons and careers.
"Right now, there are so many different silos," Beachum said. "There are so many different platforms, so many different use cases, but there's nothing that pulls it all together."
Data management platforms from tech giants such as AWS and Microsoft attempt to make it easier to integrate data from different sources, as do tools from data platform vendors Databricks and Snowflake. But when unstructured data is involved, even though techniques such as vectorization simplify discovering information contained in text and images, integrating data from multiple sources is exponentially more difficult than when dealing with structured data alone.
If AI has the ability to take all this unstructured data and pull it together, that not only helps teams from an evaluation standpoint but also helps players from a human performance standpoint. That's the Holy Grail.
Kelvin BeachumOffensive tackle, Arizona Cardinals
Therefore, if AI can help teams cull all their data and present it to coaches and players in a simple, easily digestible way, that would represent a breakthrough, according to Beachum.
"If AI has the ability to take all this unstructured data and pull it together, that not only helps teams from an evaluation standpoint but also helps players from a human performance standpoint," he said. "That's the Holy Grail that still hasn't been figured out, and which many people are trying to figure out."
The practical future of analytics in football, therefore, is that unified view of data, Beachum continued. He noted that the technology to integrate data exists in platforms from AWS, Google Cloud, Databricks and Snowflake, but the NFL has not embraced it yet.
"A platform like Databricks has done a phenomenal job, but the NFL has not gotten to the point where they can bring all these functions into a platform that provides insights," he said. "It would be ideal if we could get it done within the next decade. You're pushing the time frame if we can get it done by 2030."
Like the NFL, analytics in Major League Soccer before GenAI was somewhat primitive compared with leagues such as Major League Baseball and the National Basketball Association that had teams among the first to embrace data-driven decision-making. It was even primitive compared with other soccer leagues such as England's Premier League and Spain's LaLiga.
MLS teams don't have the financial strength of teams in soccer's top international leagues, nor do they have the same wealth as franchises in the four main U.S. sports.
AI, however, is enabling teams such as the Philadelphia Union to do far more with data than it could just a few years ago, according to Addison Hunsicker, the Union's senior manager of soccer analytics. Now, the club -- which like MLB's Texas Rangers and Minnesota Twins and the WNBA's Indiana Fever is a Databricks customer -- uses AI to create new metrics, evaluate players and even build chatbots that team personnel can use to get immediate insights on topics such as salary cap rules.
"Predictive analytics and the query piece," Hunsicker said, regarding what AI has enabled the Union to do with data that it couldn't in the past. "[AI helps us] get the most out of tracking data that we're getting on each MLS team and other leagues abroad, and build customized metrics based on our style of play."
The team's plans for further use of AI are similarly practical, Hunsicker continued.
Just as the model that the Union built to analyze players from 45 different leagues is geared toward constructing the optimal roster and learning about potential opponents, future applications of AI are focused on making it easier and more efficient to glean insights from data. One project toward that end that the team aims to tackle with AI is streamlining the delivery of post-match reports and scouting reports on upcoming opponents, which are still sent in PDFs.
In some ways, the biggest challenge for the Union isn't building new AI tools. Instead, with limited resources -- including the absence of AI experts, it's deciding what to prioritize.
"Once we figure out the use cases we want to tackle, they've become easier to get up and running," Hunsicker said. "But it's still a challenge for someone like me whose background is in data science and not AI and machine learning."
With AI enabling more widespread use of analytics in sports, football, which features more complex matchups than other sports, could be transformed the most by emerging technology.
A transformative future
Despite the focus on practical improvements, there are more spectacular ways AI will change analytics in sports in the coming years, according to UNH's Zaimes.
The NFL has perhaps the most room to benefit from AI, he said.
In sports such as baseball, basketball, hockey and soccer, most of the focus is on the action around the ball. Off-the-ball positioning and movement are important, but they're also easy to track and analyze with cameras.
In football, because there are so many matchups that go into each play, and because there are so many different play-calling possibilities, analytics is perhaps most difficult to apply to the sport. Strategies such as scenario planning to help determine whether to run or pass on a particular down were always possible. But analyzing all the individual matchups, including potential player substitutions, to make real-time decisions such as calling the perfect play using the right personnel group was more onerous.
"Because football has more scenarios than a sport like baseball, there are more combinations of things that can happen, so there's a lot more room to grow from AI," Zaimes said. "Baseball was out in front [with analytics] and is still doing a lot of things, but it has hit a bit of a plateau. There are only so many combinations with one person at bat, one person pitching and eight other players in the field."
In the near term, AI will better inform the cat-and-mouse game between offenses and defenses, he continued. For example, teams will be able to analyze how their win probability changes based on a particular play call against a specific defensive alignment or how using a given personnel grouping for a single play affects the likelihood of winning.
"That's rich, that's game-changing," Zaimes said.
Even more rich are some of the longer-term possibilities.
Assuming AI can help the teams better integrate all their data, as AI development evolves beyond reactionary chatbots to agents capable of acting autonomously to take on tasks and generate insights, AI will enable teams to design new plays and use new player combinations in the middle of a game, according to Zaimes.
For example, after analyzing the game's action, an agent might design an offensive formation a football team has never used before and come up with a brand-new play. Similarly, based on what it observes in an opponent, an agent might suggest a lineup a basketball coach has never thought of or a pitch sequence a catcher didn't consider.
"The novel element is that [decisions will be made] based on real-time tendencies that are on a tablet and shown to coaches," Zaimes said. "It's novel play-calling. That's coming, and I can't even wrap my head around what that's going to look like."
Beyond real-time creativity, virtual reality could become a means of helping players and teams prepare, according to Beachum.
Such technology, although not there yet, could enable players to scout opponents not by watching them on film but by giving them a view from inside the action. In football, quarterbacks could see simulated plays from their perspective, linemen from their perspective, defensive backs from their perspective. And the same holds true for other sports, with players at each position providing views from their unique vantage point.
"Scenario planning from a player's perspective looks like AR/VR, but we're not yet there in a way that's appetizing to a [player]," Beachum said. "We're still a ways away from that use case at the moment."
In all sports, the teams that ultimately use AI most wisely to improve analytics will be the ones that succeed, according to Alexander Booth, a solutions architect at Databricks and formerly assistant director of research and development for the Texas Rangers, including in 2023 when they won the World Series.
Wise use of analytics and AI in sports, meanwhile, does not mean the most extensive or most advanced, Booth noted. Instead, it includes a human element, one that fosters an environment of curiosity around data rather than resentment and values experience and intuition.
"We will see AI agents working like assistants, preparing reports, highlighting videos or calling out risks in real time," Booth said. "[But] the teams that succeed will be the ones that combine these tools with human intuition and a culture that embraces new ways of learning. In the long run, this blend of people and AI will change how every part of the game is played, coached and experienced by fans."
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.
