Prediction Markets Are Projected to Out-Trade Sportsbooks This NFL Season. The Two Numbers Do Not Measure the Same Thing.

Polymarket and Kalshi are projected to trade $36.8bn on NFL outcomes this season, more than double the roughly $16.75bn they handled last year. Legal US sportsbooks are projected to take about $32.3bn in NFL wagers over the same season. Put the two side by side, as most of the coverage since 18 September has, and the conclusion writes itself: prediction markets are about to overtake sportsbooks.

It does not follow, because the two figures are not denominated in the same thing. Exchange volume and sportsbook handle count different events, and the difference between them is not a rounding adjustment. It is a multiple, and nobody has published the multiple.

What volume counts that handle does not

Take a single contract with $100 of notional value. On an exchange, that contract is matched between a buyer and a seller, and the matched notional is recorded as volume. The position can then be sold before the game settles. Someone buys at 40 cents, sells at 55, the new holder sells again at 60 before kickoff. Each of those transactions is matched notional and each is recorded. One underlying $100 exposure can contribute several hundred dollars of volume across a week, and in a liquid market on a nationally televised game it routinely does.

Sportsbook handle counts the stake once. A bettor puts $100 on a side, the book takes the other side, and the figure reported to the state regulator is $100. There is no secondary market and no exit before settlement, so there is nothing to count twice. A bettor who wants out cannot resell the ticket; at most they can place an offsetting wager, which the book records as new handle, but that is a fresh $100 of exposure rather than a transfer of the original.

So the ratio between volume and handle for economically identical activity depends entirely on turnover. And turnover is the one variable an exchange is built to maximise and a sportsbook structurally cannot have. Kalshi and Polymarket want positions to change hands; that is where the fee revenue is. The comparison therefore rewards the exchange model on a metric the sportsbook model is not even playing.

Here is the limitation that should be stated plainly rather than buried: this analysis cannot compute the multiple either. Neither exchange publishes an average holding period, a resale ratio, or a unique-position count for NFL contracts. Without one of those, nobody can say what $36.8bn of volume corresponds to in handle terms. That includes everyone currently reporting the crossover. The honest statement is not that prediction markets are smaller than the headline suggests; it is that the headline compares two numbers whose conversion factor is unknown and greater than one.

The forecast spreads are wider than the gap

There is a second problem, and it is arithmetic rather than definitional. The sportsbook projection is not a single number in circulation. It is reported as $32.3bn in some coverage and $29.5bn in other coverage from the same week. That is a $2.8bn spread on one side of the comparison.

The prediction-market projection is not a single number either. The $36.8bn figure is the one being quoted, but at least one industry analysis published this month puts full-season prediction-market NFL trading volume as high as $57bn. That is a $20bn spread.

Set those against the margin being described as decisive. $36.8bn against $32.3bn is a gap of $4.5bn. It is smaller than the disagreement within the sportsbook forecast plus the disagreement within the prediction-market forecast, and dramatically smaller than the latter on its own. A $4.5bn edge derived from two estimates that individually range by $2.8bn and $20bn is not a finding. It is noise with a headline attached.

All four of these figures are projections for a season that is three weeks old. None is settled data.

What Week 1 actually shows

The hard numbers available are the Week 1 actuals, and they are more interesting than the forecasts.

• Kalshi and Polymarket combined passed $1bn in NFL trading activity over the first six days of Week 1.
• Kalshi's NFL volume over that period was $983.4m, according to industry tracker DeFi Rate, far ahead of every other venue.
• Total prediction-market volume across all exchanges in Week 1 has been reported at approximately $3.12bn.
• Kalshi held roughly 76% of sports and parlays volume, with Polymarket and DKeX the next largest.

The concentration is the part the sector-level framing hides. This is not prediction markets as a category closing on sportsbooks. It is one venue, Kalshi, holding roughly three quarters of the activity, with everything else spread across at least seven other exchanges. Any claim about the category is substantially a claim about a single company.

There is also an asymmetry in where these numbers come from that matters more than the gap between them. Sportsbook handle is reported to state gaming regulators under a statutory definition with audit exposure. Exchange volume is self-reported or compiled by industry trackers such as DeFi Rate and Prediction News, using definitions the exchanges themselves set, at companies with a direct commercial interest in the number being large. One side of this comparison is a regulatory filing. The other is a disclosure. Treating them as comparable evidence is a choice, and it is not a neutral one.

The channel doing most of the work

Growth in prediction-market sports volume is attributed in significant part to users in states where sports betting is unavailable or restricted. California, Texas and Georgia have no legal sportsbooks. Florida's market operates under the Seminole Tribe's Hard Rock exclusivity.

This breaks the share-shift framing at its foundation. A user in Texas trading NFL contracts on Kalshi is not a bettor who left DraftKings. There was no DraftKings to leave. Volume originating in a state with no legal sportsbook cannot be handle that a sportsbook lost, and counting it in a crossover comparison double-counts the market: it treats newly accessible demand as though it were captured demand.

What would make the crossover claim testable is a comparison restricted to states with mature legal sportsbook markets, measured on a turnover-adjusted basis so that exchange volume is converted to unique positions before being set against handle. No published dataset does this. Until one does, the crossover claim is not falsifiable, which is a different and worse problem than being false.

What survives

Quite a lot, actually, and it does not need the overtake framing. Prediction-market NFL volume is projected to more than double year over year, from roughly $16.75bn to $36.8bn. Week 1 produced the largest week the sector has recorded. One venue has built a dominant position inside it in under two seasons, and a meaningful share of that growth comes from demand that the legal sportsbook industry has never been permitted to serve.

That is a genuine structural shift in American sports wagering and it is worth taking seriously. What is not established, by these numbers or by any currently published version of them, is that prediction markets have overtaken sportsbooks. They would not establish it even if both projections landed exactly, because the two figures were never measuring the same quantity.

Sources: CNBC, 18 September 2026; iGaming Business; Prediction News; DeFi Rate; Bonus.com. Projection figures are analyst estimates, not settled results. Exchange volume figures are industry-tracker compilations of self-reported data.