Skip to content
Become a Partner

How a Sportsbook Makes Money: Hold, Margin and Why Parlays Drive Up to Two-Thirds of US Revenue

The main sources of revenue for bookmakers and the role of parlay bets in their revenue

A bookmaker doesn’t make money off players’ losses. It makes money off the overround built into the odds and from a mix of bet types: a single bet leaves the bookmaker 4–5% of handle, a parlay 13–30%. Using financial reports from BetMGM, Flutter, DraftKings, and Kambi, along with data from New Jersey, Illinois, and Maryland, we’ll break down how $100 million in handle is transformed into the money used to pay taxes and affiliates.

Key Takeaways

  • On a single bet with odds of −110/−110, the bookmaker builds in about 4.5% of handle (an overround of 4.76%). On a parlay with the same outcomes, the margin compounds: three events yield about 13%, and five yield about 21% (based on a 15M calculation). In New Jersey in January 2026, parlays held 21% and generated 62.5% of the state’s revenue on about a third of handle;
  • There are two filters between handle and cash. At BetMGM in Q3 2026, the GGR hold was 10.7%, and the NGR hold was 6.7%: bonuses and promotions ate up 4 percentage points of handle, or 37% of gross revenue. Next comes state tax: ranging from 6.75% in Nevada to 51% in New York;
  • The actual hold can deviate from the structural hold by 2–3 percentage points. In September 2025, Illinois’s handle increased by 9%, while revenue fell by 26%. For affiliates on RevShare, this is why monthly payouts do not correspond to traffic.

Glossary: Vig, Overround, and the Three Types of Hold

Let’s start with a classic pair. For the total or the spread, an American bookmaker sets the odds at −110 for both sides: to win $100, you need to bet $110. The implied probability of each outcome is 110/210 = 52.38%. The sum of the probabilities for both outcomes is 104.76%. That extra 4.76% is the overround, also known as the vig (vigorish) or juice: the commission built into the price.

Calculation by 15M No. 1. If $110 is wagered on each side, the bookmaker has collected $220 and will pay out $210 in either case. The remaining $10 represents 4.55% of the handle. This is the theoretical hold: the amount the bookmaker retains when the line is perfectly balanced. As a percentage of the payout, this is 4.76%; as a percentage of the handle, it is 4.55%. Operators and regulators calculate the hold based on the handle, so we’ll use the second figure going forward.

Actual hold is what you get in practice: revenue divided by handle for the period. The line is never perfectly balanced; favorites win in streaks, and the public bets on the same outcomes, so the actual hold fluctuates around the theoretical hold. Flutter introduced a third term: structural hold — the expected hold under normal sports results, taking into account the product mix. This is the figure Flutter reports to investors; Flutter and DraftKings attribute deviations from it to “sports results.”

Important: Hold is not the “percentage of losing bets.” With a 10% hold, nine out of every ten dollars are returned to players as winnings. The bookmaker makes money on the difference between the probability it offers and the probability that actually materializes, multiplied by the handle.


Handle → GGR → NGR: Three Numbers That Cause Confusion

The difference between the total volume of bets, gross revenue, and net revenue for a bookmaker

Handle – the full amount wagered, including all bets made with bonus funds (BetMGM says it includes both cash and bonus bets in its definition of handle). Gross Gaming Revenue (GGR) – Handle less the amount won by customers. Net Gaming Revenue (NGR) – GGR less promotional expenditures (e.g., free bets and odds boosts, deposit bonuses, etc.). In most states, a tax is imposed on GGR (sometimes with a partial promo deduction), while affiliate RevShare is most often calculated based on Net Gaming Revenue and after taxes.

In its Q3 2026 update, BetMGM reported the full chain. Online handle was $3.24 billion (+3% year-over-year), GGR hold was 10.7% (10.1% a year earlier), and NGR hold was 6.7% (6.4%). The 4-percentage-point gap, measured against handle, is attributable to promotions. For the first nine months of 2026, GGR hold was 9.8%, and NGR hold was 5.9%. This means that for every dollar of gross revenue, approximately 37–40 cents is returned to players in the form of bonuses, even before taxes.

“BetMGM delivered a strong Q3 as we continue to execute with discipline and generate profitable growth,” said Adam Greenblatt, CEO of BetMGM, in the Q3 2026 business update.


Why a Parlay Holds 13–30% While a Single Bet Holds 4–5%

The mechanics are simple: the odds multiply, and the margin built into each leg compounds with them. Let’s take a parlay of several outcomes at −110, each with a “fair” probability of 50%.

Calculation by 15M No. 2. The decimal odds of −110 are equal to 1.909. A parlay of three such outcomes pays 1.909³ = 6.96, whereas the fair odds for three coin flips are 8.00 (2³). The expected payout per dollar is 6.96/8 = 0.87, meaning a hold of 13%. Four legs: 1.909⁴ = 13.28 vs. 16 — hold 17%. Five legs: 25.4 vs. 32 — hold 21%. Each leg adds about 4 percentage points to the bookmaker’s edge, and this does not take into account the fact that in parlays, players more often bet on favorites, props, and “leading” outcomes with a wider overround.

A same-game parlay (SGP) adds a second layer. Outcomes within a single game are correlated: if a quarterback gains 300 yards, his team is more likely to go over the total. The SGP engine evaluates this correlation and reduces the odds more significantly than a simple multiplication, which is why the SGP margin is higher than that of a standard parlay. In the second quarter of 2026, Flutter attributed the growth of its structural GGR margin to 14.0% to “high penetration of Same Game Parlays during the World Cup.” During the same World Cup, Kambi recorded an operator trading margin of 18% with a handle exceeding €1 billion, while the share of Bet Builder in live betting among its partners rose to 22%, up from 3% in 2022.


What the State Data Show

Regulators in New Jersey, Illinois, Maryland, and Nevada publish revenue by bet type, while New York does not. We’ve compiled the published figures into a single table.

Bet TypeStatePeriodHold / ShareSource
Singles, −110/−110TheoryAny period4.5%Calculation by 15M No. 1
Singles by sportNew JerseyJanuary–September 2022Baseball 4.4%; football 3.3%; basketball 1.7%Covers according to the NJ DGE report
ParlayNew JerseyJanuary–September 202217.4%Covers according to the NJ DGE report
ParlayNew JerseyJanuary 202621% ($71.4 million on $340.1 million); 62.5% of the state’s revenueNJ DGE via InGame
ParlayNew JerseyMarch 202614.3% ($46.3 million on $323.6 million); 52.9% of revenueNJ DGE via InGame, calculation of the share by 15M
ParlayIllinoisDecember 202521.9% ($100.8 million on $460.6 million) with a total hold of 10.6%IGB via InGame
ParlayIllinoisSeptember 202513% with an overall hold of 7.1%IGB via Covers
ParlayIllinoisFiscal Year 2026~62% of players’ losses ($879 million), 3 out of 5 betsCOGFA «Wagering in Illinois 2026»
ParlayMaryland202567% of revenue with 36% handleMaryland Lottery and Gaming via CasinoBeats
Parlay cards vs all other betsNevadaSince 198430.97% versus 5.6%UNLV Center for Gaming Research via Covers
3/4/5-leg parlay at −110Theory—13% / 17% / 21%Calculation by 15M No. 2

Table 1. Hold by bet type and the share of parlays in revenue. Sources: 12–18; “Calculation by 15M” rows — see methodology at the end.

The picture is consistent: parlays account for about a third of the handle and generate between half and two-thirds of the revenue. The hold on parlays ranges from 13% to 22% month-to-month, while on single bets it ranges from 2% to 5%. Therefore, a month in which NFL favorites win one after another hits parlays the hardest: in September 2025, Illinois’ parlays held 13%, the overall hold rate fell to 7.1%, and the state’s revenue declined by 26% while handle increased by 9%.


What Operators and Providers Say

Flutter (FanDuel) discloses its structural margin directly. In Q4 2025, the U.S. sportsbook’s structural GGR margin was 15.5% (+90 bps year-over-year, driven by an 80-bps increase in the parlay share of handle), while the net revenue margin was 8.9%. In Q2 2026, the structural margin was 14.0% (+40 bps), and the net revenue margin was 8.7% compared to 10.4% a year earlier; sportsbook revenue fell 15% to $1.04 billion on handle of $11.96 billion, and the company attributed approximately $21 million of the decline to unfavorable results, below the expected level. FanDuel’s share of U.S. sportsbook GGR is 39%.

DraftKings doesn’t reveal a structural hold, but it shares a net revenue margin for its sportsbook: that’s the sportsbook revenue divided by the total handle. In the first quarter of 2026, it reached 7.8% (+140 bps) with a handle of $14.1 billion and revenue totaling $1.1 billion; the parlay share of the handle increased by nearly 300 bps. In the second quarter of 2026, the margin dropped to 6.8% from 8.7% the previous year; management stated that game outcomes led to a loss of roughly $80 million in revenue.

A B2B provider has its own metric. In Q2 2026, Kambi reported a record operator trading margin of 14.0% across its network of operator clients and expects it to normalize to about 11% for the rest of the year; its Q3 report is due on November 4. On October 22, Betsson reported a sportsbook margin after free bets of 8.1%, versus 8.8% a year earlier and 10.5% a quarter earlier, on handle of €1.37 billion; the eight-quarter average is 9.0%. A European mix of soccer and single bets yields an 8–10% margin; a US mix with parlays yields 10–15% gross margin. This is not a difference in trader skill; it is a difference in product.


Promotions and Taxes: The Waterfall From $100 Million in Handle

Distribution of a bookmaker's revenue among taxes, bonuses, and other expenses

Now let’s assemble the entire chain. Take a hypothetical online operator with $100 million in monthly handle, a structural GGR hold of 10%, and promotions at BetMGM’s level (4 percentage points of handle). Tax is calculated on GGR at state rates as of October 2026: New Jersey: 19.75%. Pennsylvania: 36% (34% to the state plus a 2% local share assessment). New York: 51%. Illinois: a progressive 20–40% scale plus a per-wager fee of $0.25 on each of the first 20 million wagers in the fiscal year and $0.50 after that. Let’s assume Illinois’s 40% rate is applicable (paid by FanDuel and DraftKings), and 3 million bets are placed at $0.50 each.

StepNew JerseyPennsylvaniaIllinoisNew York
Handle$100.0$100.0$100.0$100.0
Payouts to Players (90%)−$90.0−$90.0−$90.0−$90.0
GGR (hold 10%)$10.0$10.0$10.0$10.0
Promotions and Bonuses (4% of handle)−$4.0−$4.0−$4.0−$4.0
NGR (hold 6%)$6.0$6.0$6.0$6.0
State tax on GGR19.75% → −$1.9836% → −$3.6040% → −$4.00 + $1.5 per-wager51% → −$5.10
Remainder after taxes$4.02$2.40$0.50$0.90
Remainder, % of handle4.0%2.4%0.5%0.9%

Table 2. Waterfall: handle → GGR → NGR → after tax, per $100 million of handle, in millions of dollars. Calculation by 15M based on assumptions from the text; tax rates — sources 20–22; hold and promotions — based on BetMGM Q3 2026 (source 1).

From this remaining amount, the operator also pays platform and data providers (usually a percentage of NGR), payment processing fees, salaries, and marketing expenses, including affiliate commissions. In New York and Illinois, with a 10% hold, less than a cent of every dollar of handle is left before marketing costs. This explains why operators in those states were the first to introduce markups on odds and cut back on promotions, and why shifting the mix toward parlays is not a whim for them but a matter of survival: with a 13% hold instead of 10%, the remainder in New York rises from $0.9 million to $2.4 million on the same $100 million of handle (based on a 15M calculation).

The scale of this is evident when looking at New York: for fiscal year 2024/25, the state’s handle totaled $23.94 billion, with GGR at $2.14 billion — meaning an average hold of 8.9% — and more than half of that revenue went to the state budget.


Volatility: When Hold Deviates From Structural Hold

The structural hold represents expectations, while the actual hold reflects reality, and the difference between the two over a quarter runs from tens of millions to over $100 million. In Q4 2025, Flutter estimated the impact of these results at −$140 million in revenue and −$100 million in adjusted EBITDA, even though the same factor had been positive a year earlier. DraftKings lost approximately $80 million in the second quarter of 2026. Betsson attributed the drop in its margin to 8.1% to “player-friendly results.”

At the state level, here’s how it looks. New Jersey in September 2026: sports betting revenue of $77.0 million, down 14.3% from September 2025; online revenue of $72.9 million, down 18.8%; year-to-date revenue of $748.4 million, down 6.3%. Illinois in June 2026: handle +20% thanks to the World Cup, hold 8%, revenue −21.2%. Traffic was up, but revenue was down.

For affiliates, this means that RevShare on sports betting is a bet on hold, not on traffic. A single month of NFL action, when the favorites win, can wipe out the commission from players who lost last week — and for operators with negative carryover, it can even drag down the following month.


What This Means

To affiliates. Compare your affiliate program report to the state’s publicly available reported hold. If the NJ DGE reported a 7% hold for the month and your RevShare report shows zero, it is normal. If the report from the state shows a 12% hold, and your report shows zero payout, look into it further. BetMGM’s Q3 NGR hold after promos (but before tax) is 6.7% of handle, not 10.7%. With a 36–51% state tax rate, the same RevShare percentage applied to an after-promo, after-tax base (per Table 2: $2.40, $0.90 and $0.50 on $10.0 of GGR) pays 4–20 times less than on GGR. For sports betting, a CPA + RevShare hybrid, or RevShare with a minimum guarantee, better protects your interest in months of low hold compared to a high percentage rate.

To operators. A mix of bets is the main driver of margin, which doesn’t require raising the overround on single bets or sacrificing sharps. Flutter added 90 basis points to its structural margin in Q4 2025, driven by an 80-basis-point increase in the share of parlays; DraftKings increased its share of parlays by 300 basis points over the past year. Increasing the share of parlays increases the variability of the results. Weekends with heavy favorites in the NFL hit SGP bets harder than totals. The structural hold of sportsbooks is trending up, and so is the variability of results around that hold.

To providers. The margin is the product. Kambi doesn’t sell a trading feed. It sells an “18% margin on the World Cup with over €1 billion in handle” and a product in which 75% of bets are assessed by AI. The SGP engine, correlation model, live bet builder, and the broad array of props are all part of the product purchased. “Our hold is 1 percentage point higher” is equivalent to $30 million in GGR with a $3 billion quarterly handle.

Methodology

As of November 4, 2026. Amounts shown are in U.S. dollars, except for Kambi and Betsson (shown in euros). Hold is shown as a percentage of the handle and is calculated as revenue divided by the handle. Calculation by 15M No. 1: For the overround for the pair, −110/−110 = 2 × 110/210 − 1 = 4.76%; hold as a percentage of the handle = 1 − 1/1.0476 = 4.55%. Calculation by 15M No. 2: Hold for an n-leg parlay with outcomes of −110 and fair probability of 50% = 1 − 1.909ⁿ/2ⁿ; other factors such as correlation and varying width of overround in different markets are ignored; thus, the actual hold is likely greater than the value calculated. Table 2: Other assumptions are discussed in the text; for the estimate of tax, GGR is used, and tax is applied to GGR with no promo deduction, which is conservative for some jurisdictions; Illinois’ per-wager tax = 3,000,000 wagers × $0.50; the parlay share of New Jersey revenue for March 2026 and BetMGM promo as a percentage of GGR are 15M calculations from published figures. Data for the last 12 months from Legal Sports Report (NJ 32%/65%, MD 36%/63%, IL 31%/61%) are not shown: the webpage is unavailable.

Sources

Author of the article

Content writer

John is an experienced content strategist. He has nearly a decade of experience in creating highly valuable online content that informs and inspires. John is a co-founder of 15M, and the Managing Partner of Belianin.

Author Profile

Have you enjoyed the article?