A single percentage can describe either a pricing edge or a month’s actual outcome.
A sportsbook may post a 9% monthly hold even though its odds appear to contain only a 4%–5% margin. How can both figures be accurate? They measure different things.
Top Offshore Sportsbook Bonuses for September 2026
$2,500 Use code - PROMOBUCK2 to get a 125% First Deposit Bonus at the BetNow sportsbook when you deposit with Crypto.
$2,750
200% Crypto Bonus PLUS 10% Gamblers Insurance with a minimum $100 deposit and a maximum of $1,500. Use Code: JOIN200 in the Cashier.
Observed hold is actual sportsbook revenue divided by the amount wagered, usually over a stated period. It rises or falls with game results, bettor performance, promotions, and when wagers settle. Theoretical hold is the advantage embedded in the prices before any result occurs—often estimated from the implied probabilities on all sides of a market. Across the wider sports betting industry, both are commonly shortened to “hold,” making comparisons easy to misread. A strong month does not necessarily mean the odds became less competitive.
Start with the right figures
Before calculating hold, confirm what each figure includes. Handle is generally the total amount staked during a reporting period, but reports may treat voided wagers, refunded stakes, free bets, or bets placed in one month and settled in another differently. The difference between betting handle and revenue therefore depends partly on the operator’s or regulator’s definitions.
Gross gaming revenue (GGR) is usually handle minus winnings paid to bettors on settled wagers:
Realized hold = GGR ÷ handle × 100
“Gross” does not always mean untouched. Some jurisdictions exclude promotional credits or voids before publishing GGR, while others report adjustments separately; taxes, operating costs, and supplier fees are normally not deducted. Reporting notes matter as much as the headline number.
Before using any percentage, check its label and source:
- Realized hold: Based on actual handle and settled results for a stated period.
- Odds-derived margin: Estimated from posted prices, often called vig, overround, or theoretical hold.
Only realized hold belongs in the formula above. An odds-derived margin describes pricing, not actual revenue, so multiplying it by handle produces a rough expectation—not reported GGR.
Calculate realized hold
The basic realized-hold formula is:
Realized hold % = sportsbook revenue ÷ handle × 100
Revenue is the amount left after settled payouts are deducted from accepted stakes. Using a $1 million handle and $930,000 in payouts:
- Handle: $1,000,000
- Payouts: $930,000
- Revenue: $1,000,000 − $930,000 = $70,000
- Realized hold: $70,000 ÷ $1,000,000 × 100 = 7%
In practical terms, the sportsbook retained seven cents of every dollar wagered during that period. This is a backward-looking result, not a guarantee that its odds were designed with a 7% margin.
Hold is not automatically floored at zero. If $1 million in handle produces $1.03 million in payouts, revenue is −$30,000 and realized hold is −3%. Reporting 0% would hide the loss and distort comparisons or averages.
Build a like-for-like monthly calculation
-
Fix the reporting window
Use the same opening and closing dates for both handle and GGR. Calendar months and four-week accounting periods should not be mixed.
-
Match the betting scope
Keep jurisdiction, sportsbook, channel, and sport coverage identical in both figures. Exclude any category that appears on only one side.
-
Use settled activity
Include wagers according to the report’s settlement date or other stated recognition rule. Do not combine accepted stakes with revenue drawn from a different settlement basis.
-
Apply adjustments consistently
Follow the publisher’s treatment of voids, refunds, free bets, promotional credits, and cash-outs. Record exclusions rather than silently estimating them.
-
Run a reconciliation check
Confirm that reported GGR can be reproduced from stakes, payouts, and permitted adjustments. Then divide GGR by the matching handle and multiply by 100.
Keep the source report and methodology note with the calculation so later comparisons use the same basis.
Two reports covering the same dates may recognize activity differently. One may deduct free-bet stakes or carry unsettled wagers forward, while another may not. The resulting hold percentages should not be compared until those accounting differences are resolved.
Convert odds to implied probability
American odds can be translated into the probability implied by the price. For negative odds, use:
Implied probability = |odds| ÷ (|odds| + 100)
Consider a two-sided spread market priced at -110 on each side:
- Side A: 110 ÷ (110 + 100) = 52.38%
- Side B: 110 ÷ (110 + 100) = 52.38%
- Combined implied probability: 104.76%
A fair two-outcome market would total 100%. The excess is the overround:
104.76% − 100% = 4.76%
That extra 4.76 percentage points represents the margin embedded in the quoted prices. It is useful for comparing market pricing, including how opening lines are created, but it should not be treated as a guaranteed realized hold.
Actual results depend on how stakes are distributed and which side wins. Some sources also convert overround into a margin by dividing 4.76% by 104.76%, producing about 4.55%. Stating the formula avoids confusing these two conventions.
Why overround is not expected hold
Overround measures the excess implied probability in a set of prices. It describes price structure, not the percentage of handle a sportsbook will retain. Even a two-way -110 market with 4.76% overround returns about 4.55% of equally split handle before adjustments; the percentages use different denominators.
Actual retention also depends on how stakes are distributed and which outcome wins. If most money backs the winner, the book can lose despite displaying a margin. Promotions, limits, voids, and line movement add further variation.
The same caution applies to an election market’s quoted overround. Contracts priced at 55¢ and 50¢ total 105%, but revenue is not automatically 5% of trading volume: positions may be unbalanced, trades occur at changing prices, and fees or settlement rules affect the result. Overround is a pricing signal, not a revenue forecast.
Read short samples carefully
A weekly or monthly figure can be dominated by only a few results. If much of the handle sits on the same favorites or related markets, those outcomes may win together. One popular parlay or cluster of similar positions can turn an ordinary period into an unusually strong or weak one.
Timing matters too. Futures stakes may be collected for months, while winnings are recorded when the event settles. A championship payout can depress that week’s hold even though much of the corresponding handle arrived earlier. Likewise, one high-limit customer’s large win can outweigh thousands of smaller bets.
Longer windows—quarterly, season-to-date, or trailing 12-month—reduce this noise and make trends easier to see. Still, a longer sample is not automatically representative. An NFL-heavy sportsbook, a recreational customer base, and a sharp, high-limit operation can produce very different patterns. Review hold alongside sport mix, bet type, customer concentration, and major settlements.
Why bet mix changes hold
Two sportsbooks can offer identical prices yet report very different hold. The blended figure is effectively weighted by where the handle lands—and by which customers place it.
- Parlays and same-game parlays often carry higher built-in margins than straight bets. Correlation adjustments in same-game combinations can widen that difference.
- Futures may have wider pricing and long settlement delays, so handle and revenue can appear in different reporting periods.
- Live bets and props face rapid price changes, thinner markets, and more varied margins than standard pregame sides or totals.
- Promotions can reduce net revenue through free bets, odds boosts, or credits, although accounting treatment varies.
Customer mix matters too. Skilled bettors tend to seek stale or generous prices, while recreational bettors may favor higher-margin combinations. This helps explain the commercial rationale for limiting consistently winning bettors, though policies differ by operator.
Stake concentration also matters: one large wager can dominate results even when most tickets are small. Pricing controls, liability limits, hedging, and line movement shape the final profile. Consequently, blended hold can rise simply because more money shifted into parlays or props—not because any individual market became more expensive.
Make hold comparisons like for like
-
Match the reporting period and jurisdiction
Use identical start and end dates, and avoid combining markets with different reporting rules.
-
Align channel and product scope
Compare retail with retail and online with online. Keep sports, leagues, and pregame, live, parlay, and straight-bet mixes consistent.
-
Check settlement status
Confirm whether open wagers are excluded and whether voids, cancellations, resettlements, and cash-outs receive the same treatment.
-
Use the same revenue definition
Identify whether revenue means GGR before promotions or an adjusted figure after specific deductions.
-
Weight results by handle
Calculate combined hold as total revenue divided by total handle. A 5% hold on $900,000 and 15% on $100,000 produces 6% overall—not the 10% simple average.
Sportsbook hold measures revenue retained from stakes under a stated accounting definition. It does not show bottom-line profitability. Bonuses, gaming taxes, platform and payment fees, data costs, payroll, marketing, and other operating expenses may substantially reduce—or eliminate—that revenue.
Three questions before trusting a hold figure
- Does the numerator use clearly defined revenue, including any promotional or settlement adjustments?
- Does the denominator include the same population of accepted stakes?
- Is the period long enough—and the settled sample large and representative enough—to support the comparison?
A credible report publishes raw handle and revenue alongside the percentage. It labels realized hold separately from theoretical margin and discloses accounting differences that could change either input.
