Welcome to The Signal, Box Score's Wednesday newsletter. This week, we're looking at how AI is changing fantasy football, and how you might be able to gain an edge with AI, even when all your friends are using it, too.
While many businesses are struggling to apply AI to their workflows, fantasy football leagues have had no such issues. Adoption of AI tools in fantasy football is startlingly high. According to a 2025 Qlik survey, more than two-thirds of fantasy football participants use AI regularly to guide their teams and nearly a third would let AI manage their lineups.
But can AI actually help you win your league? The answer is complicated. For starters, fantasy football isn’t a “perfect information game” like chess or Go. In those games, nothing is hidden and there is no element of chance. The player who makes better decisions wins. These elements allow AI-powered programs like AlphaZero to become superhuman, according to Jon Krohn, author of Deep Learning Illustrated and host of the Super Data Science Podcast. AlphaZero can play millions of simulated games of chess and Go, and converge on strategies no human could ever match.
“Fantasy football is the opposite,” Krohn says. “It's built on hidden information and randomness: injuries, coaching decisions, weather, a ball bouncing off a helmet into a defender's hands. No model can predict those because the data needed to predict them don't exist.”
Fantasy players don't lack information; they are inundated with way too much of it. The landscape is saturated with rankings and mock drafts. Every week during the season brings rankings, injury reports, matchup analysis, “expert” observations, podcasts and conflicting opinions from dozens of voices. AI can synthesize that sprawling stream of advice, identify where analysts agree or disagree, and distill it into a handful of decisions relevant to your roster. It can show you where the consensus is strong.
AI’s limitations in fantasy football are similar to using AI to pick stocks, another area rife with randomness and hidden information where AI tools have struggled to make money. Another limiting factor is the sample size. A fantasy football season isn’t long enough for AI to take advantage of its decision-making prowess.
“Over a long enough horizon, a well-built model will outperform nearly everyone, the same way a card counter beats the casino over thousands of hands,” Krohn says. “But a fantasy season is 17 weeks, not thousands of hands. In a sample that small, variance dominates, and your brother-in-law who drafted on gut feel can still win the league.”
In Krohn’s view, AI will become the strongest player in the league on “expected value” over time, but the bots will remain beatable on any given Sunday and in any given season. So fantasy players should look at using AI to help them make better decisions, just like in any other endeavor, but temper their expectations for fantasy glory. Here are Krohn’s suggestions, in increasing order of sophistication:
1. Drafting
AI-projection systems estimate each player's value over a replacement-level option at his position, which tells you the round where a player is a bargain rather than a name brand. Where humans anchor on last season's highlight reels, models weigh age curves, offensive-line changes, target share, schedule, etc.
2. Lineups
The useful move is to stop thinking in terms of "who's better" and start thinking in terms of distributions. A model can tell you not only a player's projected points but the range around that projection. If you're favored in your matchup, you want your steady floor players; if you're the underdog, you want the boom-or-bust guy whose ceiling can flip the result. That’s a probability calculation humans handle badly and machines handle well.
3. Trades
AI levels an information asymmetry. A trade analyzer compares rest-of-season projections for every player in a deal, factoring in bye weeks, playoff schedules and your roster’s specific needs, so you can spot when an offer that looks generous is selling you a big name whose production has already peaked.
Of course, the real challenge in fantasy football is not letting your team dominate your every waking moment. Neither AI nor Krohn can help there. The host of the world’s most popular data science podcast stopped playing fantasy because managing his teams was taking him away from his day job.
“I haven't played fantasy sports in about a decade,” says Krohn. “I get too sucked into the competition and it eats into my productivity!” 🏈
Tell Us About Your Team
We love playing fantasy sports and we want to include fantasy in Box Score. So please share your best fantasy league stories — live draft shenanigans, dramatic comebacks, heartbreaking playoff losses — at [email protected].
If you enjoyed this edition of Box Score, pass it along to a friend and encourage them to sign up at boxscoresports.com. We’ll see you again on Friday.
