The AGA Says NFL Betting Has Stopped Growing. Its Own Numbers Say Something Narrower.

Editor's note: TrueEdge builds odds tools and earns affiliate commissions from licensed sportsbooks. This piece argues a position on prediction-market regulation, and the commercial interest runs toward the sportsbooks. Weigh it accordingly.

The American Gaming Association put a number on the 2026 NFL season this morning: $29.5 billion wagered with legal U.S. sportsbooks, against $29.4 billion last season. That is a rounding error dressed as a forecast, and it is the first time since the legal market opened in 2018 that the trade group has projected a football season without growth.

The release does not leave the reader to wonder why. "Since the widespread launch of backdoor sports betting on so-called 'prediction markets,' the growth of legal handle has stalled," AGA president Bill Miller says in it. The word "backdoor" appears in the headline of the release itself.

He has a case. He does not have the case the number implies.

Nobody agrees what last season's number was

Start with the denominator, because the entire "flat" framing depends on it. The AGA's baseline for the 2025 NFL season is $29.4 billion. RotoWire, which published its own NFL projection on August 11 and updated it today, puts last season's NFL handle at $31.76 billion — and projects $32.3 billion this year, a record.

Those are not two readings of the same figure. They are $2.4 billion apart on a season that has already been played and reported. RotoWire's methodology, per its release, works from state handle reports, by-sport breakdowns where states publish them, operator earnings calls and SEC filings. The AGA's, per its own note, applies national year-to-date growth projections to the prior season's NFL total and then uses football-specific reporting from selected states to estimate what share of wagers were on pro football.

Neither is fabricated. Both are estimates built on the same incomplete public data, because most states do not break handle out by sport. But one of them produces "the first year with no growth" and the other produces "a record," and the difference is a modelling choice, not a discovery. A trade association that spent this year lobbying against prediction markets published the first one.

The $36.8 billion is not the same kind of number

The comparison the release sets up — $29.5 billion legal, $36.8 billion on prediction markets — is the part that will travel furthest and holds up worst.

Handle is money staked. A bettor puts $100 on the Seahawks; that is $100 of handle, once. A prediction-market contract is a position that can be sold before settlement and bought again by someone else, and each of those trades counts. ESPN's David Purdum flags it in a single sentence in his story today: prediction-market volume "differs somewhat from betting handle because the contracts can change hands multiple times." RotoWire says the same thing more bluntly — a single underlying exposure can generate several units of volume.

We do not know the multiple. Nobody has published a churn ratio that would let you convert $36.8 billion of NFL contract volume into the number of dollars actually risked on NFL outcomes, and until somebody does, the two figures cannot be stacked in the same bar chart. They are in the chart anyway.

The strongest evidence against substitution comes from the same analyst

Here is the number that should have led the coverage. According to Eilers & Krejcik data cited in RotoWire's release, 69% of prediction-market sports volume originates in states with no legal sports betting at all, and 44% comes from California and Texas alone.

If prediction markets were pulling bettors out of legal sportsbooks, the damage would appear in the states where legal sportsbooks exist. Roughly seven in ten trading dollars are coming from places where a legal sportsbook is not an option. Bill Speros, RotoWire's senior betting and prediction market analyst and the author of the $36.8 billion projection the AGA's framing borrows, says it in one sentence: "Prediction markets aren't stealing NFL bettors from sportsbooks so much as reaching people sportsbooks can't legally reach yet."

The more parsimonious explanation for a flat 2026 is the boring one. Missouri is the only state to have launched legal betting since 2025. The legal market grew for seven years by adding states; it has run out of states to add. California and Texas are not close. Growth in a market that stops expanding its addressable population converges on the growth rate of the population already inside it, and 0.3% is roughly what that looks like in a mature category.

The counterargument, which is real

The best version of the AGA's position is not the handle comparison. It is this: substitution does not have to be large to be visible at the margin, and the trend line has genuinely bent. Sportsbook handle growth ran at about 4% from September through May, down from 14% in the prior comparable stretch. Prediction-market volume in August was 4.6 times August of a year ago. Those are the AGA's and ESPN's numbers, not a lobbyist's adjective, and a category growing at 4.6x while the incumbent grows at 4% will eventually take share whatever the current geographic mix says. The 69% figure is also a 2025 baseline; if the mix is shifting toward legal states, the substitution argument gets stronger every month and the trade group would be right early rather than wrong.

And the AGA's strongest point has nothing to do with money at all. It estimates $5.1 billion in sports-contract volume from users aged 18 to 20 — an age group barred from sports betting in 35 of the 40 states that have legalised it. Fortune's August 31 report, working from a CNN analysis, put 18-to-21 trading on Kalshi at $5.4 billion across all categories with $3.9 billion of it on sports. The figures do not reconcile cleanly, and we are not going to pretend they do. Both describe a hole that a licensed sportsbook is not permitted to have.

That hole is the argument. A 19-year-old cannot open a FanDuel account in New York and can open a federally regulated exchange account and trade a Chiefs contract from a dorm room, because the exchange sits under the CFTC rather than a state gaming board. Fortune reports that a Connecticut bill to raise the minimum age on those platforms to 21 did not pass. Whether any of this costs the legal industry a dollar of handle is beside the point, and by leading with handle the AGA made the weaker case with the stronger evidence sitting further down its own release.

What the NFL did instead

The league answered this question a week before the AGA did, and answered it commercially. On August 27 the NFL renewed DraftKings and FanDuel as official sportsbook partners and added Fanatics, all non-exclusive, all with access to official league data distributed by Genius Sports. It also carved prediction markets out. The league's Renie Anderson, on the sector: "That's not a space that we're considering right now commercially." Per ESPN, the NFL's data and intellectual property are excluded from sportsbook prediction-market products in states without legal betting.

That is a more honest statement of the industry's position than the handle projection. The NFL is not claiming prediction markets are stealing its bettors. It is declining to license them, which is what a rights-holder does when it wants leverage in a fight whose outcome is still open — the Third and Ninth Circuits split on state authority in April and August, and New Jersey asked the Supreme Court to resolve it on September 2.

What would change our mind

Two things, and they are checkable. If a state that publishes by-sport handle shows an outright year-over-year decline in NFL handle this season while its neighbours without legal betting show rising contract volume, the substitution story survives contact with data and we will say so. And if an exchange publishes an open-interest-to-volume ratio that makes $36.8 billion comparable to $29.5 billion, the comparison becomes legitimate and we will run it.

Until then: the legal market stopped growing because it stopped adding states, prediction markets are mostly serving people the legal market is not allowed to serve, and the genuine scandal in today's release is a $5.1 billion figure attached to teenagers that the press release used as a supporting statistic.

Watch