With accurate player-availability predictions for all active players, AI-powered decision-making is dramatically improved around three dimensions: 1) Risk management: If a productive wide-receiver is likely to get hurt, for example, a team might invest more in talented backups, to minimize drop-off in team performance during injury. 2) Training and targeted interventions: If AI suggests a player is injury-prone, teams can target that player with customized training, nutrition, or other regimens to reduce the likelihood of injury. Alternatively, a team might choose to reduce a player’s workload, also reducing risk. 3) Personnel decisions: By identifying factors that predict injury or other unavailability, teams can draft, trade for, or otherwise acquire players that they believe are more likely to be available season-long. Additionally, teams may choose to trade players for whom injury seems likely.
March 15, 2023, marked the official beginning of the National Football League’s new season. It also meant the start of the free agency period, when teams make deals with players no longer under contract with their former teams; each new contract can literally mean millions of dollars of a team’s budget.
So it’s critical to get recruiting right. Like any business, NFL executives and leaders of other professional sports teams must make decisions about how best to allocate their limited budgets, placing informed bets on the ROI they will gain from assets (players, in this case), including as related to expected performance (on and off the field), future injuries, and other factors.
But what if this year, AI could tell us how many games a player has left in their career, how many points they will score next season, or whether they will suffer a major injury in the near future?
While free agency and other recruiting mechanisms have been around for decades, how decisions about players get made is changing rapidly. Specifically, the application of AI-based technologies to massive sets of sports data is enhancing the front office’s ability to make decisions about players — whom to recruit, develop, bench, or trade. And it will change the workings of all pro sports, permanently.
But will AI replace the front offices of sports teams anytime soon?
While this new technology is definitely augmenting human decision-making, we don’t see it replacing general management teams in the near future — in sports or other businesses.
Amid a large, growing number of AI-based sports-focused offerings, some are aimed are helping team decision-makers with predictions about athlete injury and longevity. Knowing the likelihood of injury within a specific time frame has a large impact on recruitment, as teams would naturally aim for players expected to remain injury-free longer. Industry executives have always had some experience-based intuition for factors that go into injury, such as time and “mileage” on the field. Sometimes these predictions hold true, but often they don’t.
The difference now is that AI can back up some conventional wisdom — in the NFL for instance, a wide-receiver over 30 years of age is more prone to injury and other challenges, for example — but can also provide much more specific estimates of likelihood of injury or diminished performance, and what that means for a given player’s availability and what that might cost the team. One company, Probility AI, claims 96% accuracy in predicting which players will miss time next season. Executives can use these results to go from “I think this is probably an important factor” to “I know this is an important factor and can estimate its impact and cost with unprecedented confidence.”
AI-generated insights go well beyond existing ones or those backed by intuition. For example, Probility AI trained its injury-prediction models on data from specific NFL teams, along with other public and private data sources, to understand the impact of factors like where a given player went to college, combinations of head and assistant coaches they played under, and consequent practice and workload demands. While these nascent insights warrant further research, they show how deep AI can go in its predictive analyses.
As a result, instead of general managers trying to secure the very best wide-receiver overall, they can find the best receiver for their team, based on AI predictions of future injuries and performance. Since players typically have different predicted career lengths and performance outcomes with different coaches, field conditions, or teammates, this creates an arbitrage situation whereby player market value varies depending on which team the player plays for.
Multiple NFL teams are deploying AI technologies from Probility AI and other sources, with good reason: failing to do that would put them at a disadvantage against their AI-equipped peers. Of course, such models are also being used in other sports such as soccer and basketball to generate value, and across business sectors to enhance activities including informing decisions, boosting productivity, and serving customers better.
So as AI gains predictive capabilities across key dimensions of sports — injuries, trade timing, others — will it replace the front office?
In short, no. For now, think of AI as augmenting human decision-making. It won’t replace executives but help them make better decisions, especially in areas where human error and bias are more likely, such as basing recruiting largely on intuition and doing “what worked before.” Where the Moneyball movement of the last 20 years has been about using player statistics in a much more rigorous, systematic way, AI uses deeper learning to make even better predictions about performance.
With accurate player-availability predictions for all active players, decision-making is dramatically improved around three dimensions:
Savvy executives will also integrate injury prediction into financial decision-making. That is, AI not only generates predictions on player availability, but can input those predictions in a financial decision-making engine, enabling team leaders to create granular metrics on expected productivity per dollar spent. For instance, a running back who is predicted to play in only 50% of games in a given year becomes, functionally, twice as expensive as a similar-cost one who could play every game. By considering price paid per outcome (yards gained, tackles made, points scored, others), teams can allocate their dollars in a maximally efficient manner, optimizing on productivity for each dollar spent.
However, the technology by itself isn’t enough. While software can analyze player engine and resource allocation, sports executives’ judgment and risk tolerance must eventually choose among inevitable tradeoffs and dictate the decisions made. We share more on this in the last section.
Still, AI is an absolute game-changer in professional sports and is replacing informal or even statistics-based decision-making as the engine of a comprehensive system fueled by big data and unprecedented predictive power.
It’s easy to see how better predictions — generated by AI — would have massive impact on any business. A close analogy here would be predicting when worker performance in labor-intensive industries like construction may suffer, or when large equipment like that powering manufacturing plants or refineries would malfunction or fail and taking preventative steps before a costly incident. The approach would apply to any business with aging resources.
More broadly, predicting demand for anything from clothing to corn would enable business leaders to make better decisions about production, including as related to supply chain and other areas. Other AI-based algorithms could make predictions about competition. The list goes on, and AI already has been applied in these and other ways across sectors, helping to explain why AI startups received nearly $1.4 billion in funding in 2022.
Of course there are limitations to using predictive AI, further reinforcing the idea of augmentation versus replacement.
With regard to predicting NFL injuries, for example, while new technology can guide decisions about recruiting, trades, and how much to pay a given player, the coaching team has to think strategically about the whole-team dynamic. The AI may tell you it’s time to replace an injury-prone running-back with a player with a given profile, but an executive will have to think about how best to integrate the new recruit into the team. Total risk, after all, is spread out across all the players and their interactions. Here, too, AI is getting better at understanding the big picture of teams and its implications, starting with sports with smaller starting-team sizes, such as hockey, which puts no more than six players on the ice at a given time.
Moreover, it’s important to understand that AI-based offerings aren’t providing a definitive “answer” but making a prediction with a confidence interval around it. That interval will shrink as the technology improves, but there will always be some looseness related to prediction and that’s, again, where human judgment is critical.
In the end, AI is definitely a game-changer for sports, giving front offices and coaches unprecedented predictive power to make an increasing range of decisions with large implications for performance and returns, giving players insights to extend their careers, and keeping more player playing, which is thrilling the fans. But it’s still a story of augmentation, one where leaders, using new technologies to inform their experience-based intuition, must make strategic calls the best they can and maintain accountability for what happens on the field and the balance sheet.