H2 Says NFL Handle Falls 0.8% This Season. The Number Worth Reading Is the 4-to-6% Underneath It.

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

H2 Gambling Capital's number for the 2026-27 NFL season is $31.4 billion of US sportsbook handle, down 0.8% year over year — the first decline in nine years of legal wagering. Gross win of $3.5 billion, down 2%, on a hold of just under 11%. Add college football and the projection is $42.1 billion.

Nobody should care very much about 0.8%. It is inside anyone's forecast error, it will be revised, and a firm that publishes a directional call in September has ten months to be quietly wrong about it. The reason to read H2's note is two lines further down, where it does something the American Gaming Association would not do five days ago: it separates the demand from the theft.

H2's position is that NFL betting demand in the United States is still growing at roughly 4%. Its position is also that prediction markets will remove 4% to 6% of sportsbook handle in competitive legal states this season. Subtract the second from the first and you get a small negative, which is exactly the headline. The forecast is not really a forecast about football. It is a published substitution rate, and it is the first one we have seen from anyone with a model behind it.

That matters because the argument the entire industry has been having since August has had a hole in the middle of it. The AGA said on September 4 that NFL handle would be flat — $29.4 billion in 2025, $29.5 billion projected for 2026 — and pointed at Kalshi and Polymarket. We wrote at the time that its own numbers said something narrower than its press release did, because "flat" is consistent with a dozen explanations and the AGA had offered no way to distinguish them. Maturing states. A tax pass-through in Illinois. Promotional discipline after two years of margin expansion. Weather, even. Blaming the exchanges is a hypothesis; the AGA presented it as a finding.

H2 has now put a number on the hypothesis, and the number is checkable in principle. If prediction markets are taking 4% to 6% of handle in the states where sportsbooks are legal, then handle in those states should underperform handle in states that came online recently by roughly that margin, and it should underperform on the market types where the exchanges are deepest — sides and totals on nationally televised games — while holding up on the parlays where the exchanges are still thin. Those are testable claims. In four months, monthly state filings will either show that shape or they will not.

The awkward part is who is saying it

Here is what makes the note interesting rather than merely convenient. On April 22, H2's managing director Ed Birkin published an analysis whose title was, in effect, that prediction markets are "overhyped, and underperforming." He argued that the sector's headline volume figures are close to meaningless as a competitive measure, because volume on an exchange is not handle at a sportsbook. Citing Sporttrade COO David Huffman's conversion estimates — handle equivalent to 5% to 20% of volume in golf and futures markets, 45% to 55% in game markets on basketball, football and baseball — Birkin took the $1.6 billion of prediction-market volume reported around Super Bowl LX and estimated the volume on the game itself at $633 million, converting to a handle equivalent of about $317 million. Against a licensed sportsbook handle he put at $1.4 billion, that is a 7% share in competitive states.

Then he made the point that the operators keep declining to answer. Prediction markets convert roughly 1.1% of volume into revenue. Sportsbooks hold around 10%. On Birkin's Super Bowl arithmetic, licensed books earned about $141 million on the game and the exchanges earned about $2.4 million from their share of it — perhaps $18 million if you count every prediction-market product running that week. Kalshi processed 97 million transactions across all of 2025; Betfair, at its peak, processed more than 120 million a day.

So the same firm now forecasting the first NFL handle decline in the legal era spent the spring telling its clients that the thing causing it is smaller than it looks. Both can be true, and the reconciliation is the actual story of this season: exchanges are taking handle without taking revenue. A trader who moves $10,000 through a Kalshi contract on the Seahawks and out again has removed something from FanDuel's top line and given Kalshi about $110. Handle is a volume measure. It is the measure every state publishes, every trade group quotes and every analyst models — and it is now the measure most exposed to a competitor that does not need it to be profitable.

What the Missouri line does to the headline

H2's $31.4 billion includes Missouri's first full season, which it puts at roughly $2 billion of handle. Missouri went live on December 1, 2025, so it caught the back end of last season too — December and January, which is playoff-adjacent and not nothing. We do not know how much NFL handle Missouri booked in those two months, and without it we cannot compute the like-for-like number.

We can say what the arithmetic implies. A national total that is down 0.8% while absorbing most of a new state's first full season means the states that were already open are down by materially more than 0.8%. How much more depends entirely on Missouri's December-January figure, which nobody has published in a form we could open. That gap is worth naming rather than papering over, because the difference between "the market shrank slightly" and "the mature market shrank meaningfully and a new state hid it" is the difference between a rounding error and a trend.

The gross-win line points the same direction. Handle down 0.8% against revenue down 2% implies hold falling, not rising — which is what you would expect if the customers most likely to leave for an exchange are the ones paying the widest prices, or if books are sharpening lines and giving back margin to keep them. FanDuel spent Wednesday making its 3% injury-protection fee free for the whole of September. That is not the behaviour of an operator with pricing power.

The strongest case against all of this

The best objection is not that H2 is wrong about the direction. It is that H2 sells research to the gambling industry, and an industry currently lobbying Congress and thirty-eight attorneys general has an obvious use for a forecast that says the exchanges are eating it alive. A number that flatters the client is not evidence.

That objection is real and it is the reason to weigh the April note heavily. Birkin's April analysis was the opposite of flattering: it told the same clients that the headline volume figures everyone was citing overstate competitive threat by a factor of two or more once converted to handle, and that the exchanges' revenue base is roughly a tenth of theirs per dollar traded. A shop willing to publish that in April is not obviously running the lobby's script in September. And the second-order check holds: H2's reporting-state figures of $28.7 billion for 2025 and $28.8 billion for 2026 land within 3% of the AGA's independently produced $29.4 billion and $29.5 billion. Two models, different methods, same answer.

The other objection is simpler. Prediction-market volume hit $2.67 billion in a single day last Saturday, a three-month high, off the back of college football's first full slate — and the sector has spent close to $200 million on digital advertising through July, with more than 2,400 active ads running as of Wednesday against 970 on August 10. If that is what 4% to 6% of sportsbook handle costs to buy, the acquisition economics are brutal and the substitution rate should be read as a ceiling that money bought, not a floor that demand created.

We think that cuts both ways, and we would rather say so than pick the convenient half. Advertising at that scale buys trial, not habit. The number to watch is not this Sunday's volume. It is whether the states' October filings show handle down in a way that tracks H2's 4-to-6% and not the AGA's shrug.