New York's Books Took a Record $594.6 Million and Kept 2.85% of It

Eight mobile sportsbooks took $594,594,255.88 in wagers from New Yorkers in the week ending Sunday, September 13. They kept $16,963,049.06.

That is a hold of 2.85%, and both halves of it are remarkable. In the 233 weeks since New York's mobile market opened in January 2022, only five weeks have produced more handle. Only seven have produced a lower win rate. The state has never before had a week that was near the top of one list and the bottom of the other.

Covers reported it Thursday as the third-lowest hold of any NFL regular-season week in New York history. We pulled the full series off the Gaming Commission's own filings to check, and the claim holds. The only NFL regular-season weeks below it are the week ending October 20, 2024, at 1.91%, and the week ending December 18, 2022, at 2.78%. The five other weeks that beat it to the bottom are a different animal in each case: two Super Bowl weeks (February 2023 and February 2024), the week ending March 30, 2025, in the middle of the men's tournament, the week ending September 3, 2023, before the NFL season had started, and one week in June this year that we will come back to.

The dispersion is the story

Here is the week, operator by operator, from each company's own filing with the Commission.

Book	Handle	Gross revenue	Hold
FanDuel	$224,469,223	$8,347,252	3.72%
DraftKings	$216,744,126	$4,743,263	2.19%
Fanatics	$60,516,979	$642,600	1.06%
Caesars	$36,284,766	$1,130,590	3.12%
BetMGM	$35,436,082	$966,974	2.73%
BetRivers	$10,420,724	$608,148	5.84%
theScore Bet	$7,510,657	$420,192	5.60%
Bally Bet	$3,211,698	$104,029	3.24%

Every one of those books graded the same slate. Same favourites covering, same overs cashing, same touchdown scorers. And the two smallest operators in the state — with under $18 million in combined handle, less than a twelfth of what FanDuel alone took — had a perfectly ordinary week. BetRivers at 5.84% and theScore at 5.60% roughly doubled the market.

Meanwhile Fanatics, which took $60.5 million, kept about one cent on the dollar.

If this were pure variance, it would not sort that way. Variance does not know which book is which. It would spatter across the eight more or less at random, and the two books with the smallest samples would show the widest swings in both directions, not a tidy landing near the long-run average while the three biggest got scraped.

What sorts that way is product mix. The large books are parlay businesses on an NFL Sunday, and a parlay book's losses are correlated in a way a straight-bet book's are not. Citizens JMP's Week 1 pricing survey put DraftKings' and FanDuel's implied margin on a two-leg favourite-and-over combination at 22.0%, against roughly 4.1% on a moneyline. That enormous theoretical edge is the compensation for an enormous variance: when the favourites cover and the games go over on the same afternoon, the correlated legs across tens of thousands of tickets do not fail independently. They all land together.

By Covers' count of the weekend, eight favourites covered the spread and nine of thirteen games went over. That is the exact joint event a same-game parlay book is short.

We should say plainly what we cannot establish. The Gaming Commission does not break handle out by bet type, so we cannot prove from these filings that parlay share explains the spread. What would falsify it is simple enough to watch for: if BetRivers and theScore keep landing near 5-6% in high-scoring, chalk-heavy weeks while the big three swing between 2% and 14%, the mix explanation is doing the work. If the small books start swinging as violently, it was noise and we were reading a pattern into eight data points.

This is the second September in a row

Last year's equivalent week — the seven days ending September 7, 2025, which covered the 2025 opening weekend — produced $527.6 million of handle and $30.6 million of revenue. A 5.79% hold. Also bad, in the same month, for the same reason.

So the year-over-year comparison that has circulated since Thursday morning, revenue down 44.5%, is measured against a week that was itself well below normal. Handle over the same span is up 12.7%. New Yorkers bet more money than they have ever bet in a single week and the industry's revenue fell by nearly half. Demand is not what is moving.

And this is not the floor. On the week ending June 14 this year, New York's eight books collectively lost $48,515,978.57 on $587.6 million of handle — a hold of minus 8.26%, the only negative week in the market's history. The Knicks won their first NBA title in 53 years and the United States beat Paraguay in its World Cup opener on home soil in the same seven days. FanDuel alone paid out $22.6 million more than it took in. BetRivers made $41,379 that week, and theScore lost $13,412, which is to say: the same two books that had an ordinary week last week also had a roughly ordinary week during the worst week the state has ever seen.

Across 233 weeks, New York's weekly hold has ranged from minus 8.26% to plus 17.61%. The industry talks about hold as though it were a rate — "we hold about ten and a half" — and for a full fiscal year that is a fair description. At weekly resolution it is not a rate. It is a distribution with a fat left tail, and the tail is getting fatter as the product gets more correlated.

What it costs the state, not just the books

New York taxes mobile sports wagering revenue at 51% of gross gaming revenue, the joint-highest rate in the country, and unlike most states it does not let operators deduct promotional credits from the base. The tax lands on the number in that table.

At $16.96 million of gross revenue, the state's share of last week is roughly $8.7 million. The week before, on $44.8 million of revenue, it was roughly $22.9 million. A $14 million swing in a single week, on a line item that statute directs to elementary and secondary education, youth sports grants, property tax relief and problem-gambling services.

That is the part of this the Assembly should sit with. A 51% rate on a volatile base is a bet the state has taken without quite noticing that it has taken it. New York has structured its public finance so that when nine NFL games go over on a Sunday afternoon, the schools' line item moves. In a fiscal year that mostly washes out. In a budget cycle, "mostly" is doing a lot of work.

The best argument that none of this matters

It is the honest one and it deserves the space: a sportsbook's edge is in the price it makes, not the result it gets. Every hold number is a sample. One week of thirteen football games is a tiny sample, and the correct response to a 2.85% week is to note that the expected value did not change and wait for the next fifty-one.

That is right about the level and wrong about the shape. Nobody at FanDuel is worried about a single bad Sunday, and nobody should be. What the filings show is not that the books had a bad week — it is that they had different weeks, sorted by size, in a direction that a fair coin does not produce. The three books running the most aggressive multi-leg product took the damage. The two running the least took none.

That is not a variance story. It is the shape of a business that has moved its revenue into a product with a beautiful average and an ugly tail, and then built a tax base on top of it.

Buffalo and Detroit kick off tonight with the highest total on the Week 2 board. Watch the small books.