The 2026 NFL season is the first full-season, head-to-head fight between sportsbooks and prediction markets, and the operators are saying so out loud. On second-quarter earnings calls, executives at Penn Entertainment, BetMGM, Flutter, and DraftKings all flagged record marketing and promo spend for football season — with Penn CEO Jay Snowden predicting “very aggressive, irrational marketing spend” across the industry.
Why This Season Is Different
FanDuel, DraftKings, and Fanatics — the three largest US sportsbooks by handle — all launched their own prediction-market platforms in December 2025. Last fall, those products caught only the tail end of the NFL calendar. This September, sportsbook-linked prediction markets compete against Kalshi and Polymarket for a complete season for the first time.
The warm-up act already happened. The 2026 FIFA World Cup generated a record $20 billion in prediction-market volume, according to Jefferies — the largest single event in the sector’s history, surpassing the 2024 US presidential election. Operators watched a fully engaged global audience trade a sporting event at scale, and every one of them drew the same conclusion about the NFL.
The baseline is enormous. Last NFL season, prediction markets handled roughly $16.75 billion in NFL trades, including more than $1.6 billion on the Super Bowl across Kalshi and Polymarket combined. Kalshi’s Super Bowl LXI winner market has already cleared $59 million in volume before a meaningful snap has been played, with the Rams at 16% and the Bills and defending-champion Seahawks at 8%.
The Numbers Each Side Is Bringing
DraftKings gave the clearest look at how fast the category is scaling. CEO Jason Robins said DraftKings Predictions grew from an annualized $2.3 billion in volume in April to $11 billion by July, with more than 600,000 customers since launch — and he expects volume to “explode” when the NFL starts. Robins also claims only about 1% customer overlap between the DraftKings sportsbook and the largest prediction-market operator in legal-betting states, an argument that prediction markets are expanding the pie rather than cannibalizing it.
Flutter’s FanDuel Predicts is further behind, generating just $6 million in revenue last quarter — industry estimates put it 9 to 12 months behind DraftKings. Flutter is responding with infrastructure: it announced it is shifting FanDuel Predicts’ sports and novelty contracts from CME to Crypto.com specifically to “deliver new products at pace ahead of the NFL season start.”
Penn is the notable holdout, sitting out the prediction-market race entirely to focus on sportsbook profitability under theScore Bet.
| Operator | Product | Status entering NFL season |
|---|---|---|
| Kalshi | Native exchange | Liquidity leader; $59M+ already on Super Bowl LXI |
| Polymarket | Native exchange | Deep NFL markets; 5% sports fee coefficient |
| DraftKings | DraftKings Predictions | $11B annualized volume (July); 600K+ customers |
| FanDuel | FanDuel Predicts | $6M quarterly revenue; migrating CME to Crypto.com |
| Fanatics | Prediction product | Launched December 2025; first full season |
| Penn | None | Sitting out; sportsbook-only strategy |
What the Arms Race Means for Agent Builders
For anyone running automated strategies on the agent betting stack, this season changes three things.
First, liquidity fragments. The same NFL outcome will trade on five-plus venues with different fee schedules, tick sizes, and settlement rules. Kalshi’s taker formula runs a 7% coefficient against Polymarket’s 5% on sports, and the sportsbook-linked products price differently again. That is a cross-venue spread an agent can measure continuously — the exact workload covered in the Kalshi API guide and Polymarket API guide, and tracked at the venue level by the Vig Index.
Second, promo-driven dislocations peak in September. “Irrational marketing spend” is Snowden’s phrase for boosted odds, deposit matches, and customer-acquisition pricing that temporarily detaches quotes from fair value. Those windows are short and were historically harvested manually; agents that monitor multiple venues can harvest them systematically. The math for sizing into them is standard Kelly Criterion territory.
Third, API access becomes the moat. Kalshi and Polymarket expose full trading APIs — the reason virtually all production betting bots run there. The sportsbook-linked prediction products largely do not. If the December 2025 entrants want the liquidity that comes with algorithmic flow, they will face pressure to open programmatic access; if they stay closed, the arms race splits into a retail-promo war on one side and an agent-liquidity war on the other, with Kalshi and Polymarket winning the second by default.
The Verdict Comes Fast
Unlike the World Cup, the NFL is an 18-week regular season plus playoffs — a long enough runway to separate durable market share from expensive noise. The first data point arrives Wednesday, September 9, when the Patriots visit the defending-champion Seahawks in a Super Bowl rematch. Watch where the Week 1 volume actually clears; that, not the ad spend, will say who is winning.
For the venue-by-venue execution picture, see our Kalshi API guide and Polymarket API guide, and track measured pricing across books in the Vig Index.
Have a tip or a correction? Reach out to us.
Not financial advice. Built for builders.