A hold percentage is incomplete until the denominator is named.
“Ten percent hold” can mean the casino retained 10% of slot coin-in, 10% of table drop, or 10% of settled sportsbook wagers. Those bases measure different things, so the percentages should not be compared as though they were interchangeable prices.
One formula, several denominators
\text{Hold percentage}=\frac{\text{gaming win}}{\text{defined activity base}}\times 100The numerator is the operator’s win for the reporting definition. The denominator depends on the product and report.
| Product | Common denominator | Typical formula |
|---|---|---|
| Slots | Coin-in | Slot statistical win ÷ coin-in |
| Table games | Drop | Table win ÷ table drop |
| Sports wagering | Settled handle or settled wagers | Accrual win ÷ settled wagers |
| Other games | Write, sales, or another defined base | Win ÷ stated base |
A report may use jurisdiction-specific adjustments. Always read its definitions before reproducing the percentage.
Slot hold is based on repeated wagers
Suppose a bank of machines records:
- coin-in: $1,200,000;
- statistical win: $84,000.
\text{Actual slot hold}=\frac{84{,}000}{1{,}200{,}000}=7.00\%Coin-in includes recycled credits wagered again. A player may insert $100 and create several hundred dollars of coin-in through repeated play. This makes slot hold much closer to a wagering-price ratio than table hold based on drop.
Slot reports also compare actual hold with theoretical hold, sometimes called par. If the weighted theoretical hold is 8.20%:
\text{Hold variance}=7.00\%-8.20\%=-1.20\text{ percentage points}That difference can reflect normal volatility, jackpots, meter problems, configuration errors, promotional treatment, or an insufficient sample. It is not automatically evidence that the machine is paying incorrectly.
Table hold uses money entering the table, not every wager
Assume a blackjack pit records $100,000 in drop and $22,000 in win:
\text{Table hold}=\frac{22{,}000}{100{,}000}=22\%That does not mean blackjack carried a 22% house edge. Drop is the cash, chips, or approved value recorded as entering the table’s drop system. The same bankroll can be wagered many times before it leaves the table.
If those players generated an estimated $400,000 of total action, then win divided by action would be 5.5%. Even that realized ratio is not the mathematical house edge; short-term results and player decisions still affect it.
The distinction is:
- House edge applies to wagers under stated rules and strategy assumptions.
- Table hold applies to drop, which is an operating and accounting base.
- Churn between buy-in and final cashout can make table hold much larger than the game edge.
Sportsbook hold has a timing problem
A sportsbook may report $2,000,000 in settled wagers and $140,000 in accrual win:
\text{Sports hold}=\frac{140{,}000}{2{,}000{,}000}=7.00\%But a simple gross-revenue-to-total-handle calculation can be misleading when the handle includes future events whose payouts have not yet been settled. Accrual reports often align the numerator with wagers resolved in the same period.
Massachusetts reporting materials define revenue hold as accrual win divided by current and future wagers settled. The February 2026 sports-wagering revenue report is one current example of that denominator discipline.
Actual, theoretical, expected, and realized hold
These terms overlap but answer different questions.
| Term | Meaning |
|---|---|
| Theoretical hold | Programmed or modeled long-run retention rate |
| Expected hold | Expected result for a stated mix of games and activity |
| Actual or realized hold | Win divided by the relevant recorded base for a period |
| Weighted theoretical hold | Several theoretical rates combined according to their activity share |
For three slot groups, weighted theoretical hold is:
H_w=\frac{C_1H_1+C_2H_2+C_3H_3}{C_1+C_2+C_3}where C_i is coin-in for group i, and H_i is that group’s theoretical hold.
If $500,000 is wagered at 7%, $300,000 at 9%, and $200,000 at 12%:
H_w=\frac{500{,}000(0.07)+300{,}000(0.09)+200{,}000(0.12)}{1{,}000{,}000}=8.60\%A simple average of 7%, 9%, and 12% would be 9.33% and would overstate the blended expectation because the 12% games received the least coin-in.
Percentage and dollars answer different management questions
A higher hold percentage does not necessarily produce more win. A machine bank holding 10% on $100,000 coin-in produces $10,000, while another holding 7% on $500,000 produces $35,000. Volume can outweigh the percentage.
The same warning applies to comparisons across time. A weekend can show lower hold but much higher total win because the property handled more action. Managers normally read hold alongside coin-in, drop, handle, number of decisions, average bet, and operating hours.
Payout percentage is the complement of hold only when both use the same base and accounting treatment:
\text{Payout percentage}=100\%-\text{hold percentage}For the slot example with 7% actual hold, the corresponding payout percentage is 93% of coin-in. It would be wrong to subtract a 22% table hold from 100% and claim that blackjack returned 78% of all wagers, because table hold used drop rather than total action.
The official definitions show why context matters
Federal gaming-control definitions state that actual hold percentage is win divided by drop or coin-in and allow the calculation by table, machine, game type, day, or cumulative period. See 25 CFR 542.2.
Nevada’s slot internal-control procedures are more specific: actual slot hold equals statistical win divided by coin-in, and floor par is a coin-in-weighted theoretical hold. The Nevada slot analysis requirements also require investigation of performance and meter issues rather than blind reliance on one percentage.
Five questions to ask before using a hold figure
- What is the denominator?
- Is the numerator actual, theoretical, taxable, or adjusted win?
- Does the period align settled wagers and payouts?
- Is the sample large enough for the product’s volatility?
- Were promotions, jackpots, fills, credits, or meter corrections treated consistently?
Negative hold is possible when players win more than the operator during the measured period. Extremely high positive hold can also be a short-term result. Neither should be mistaken for a permanent change in game mathematics without further evidence.
Use expected hold for the modeled result, realized hold for the observed result, and payout percentage only when the return and hold use the same wagering base. The denominator is what makes the percentage honest.