Hold percentage tells you how much gaming win remains after comparing win with a defined activity base. The percentage is useful only when that base is named. A 10% slot hold, a 10% table hold, and a 10% sportsbook hold can all be calculated correctly while describing very different economics.
The mistake is to treat “hold” as if it were one universal casino price. It is not. The numerator is usually some form of gaming win. The denominator may be coin-in, table drop, settled sports wagers, write, or another jurisdiction-defined measure. Before comparing two hold percentages, first ask whether the denominators measure the same thing.
Start with the denominator, not the percentage
\text{Hold percentage}=\frac{\text{gaming win}}{\text{defined activity base}}\times100Three terms need to be fixed before the result has meaning:
- Gaming win is the casino-side result under the accounting definition being used.
- Activity base is the denominator: for example, slot coin-in or table drop.
- Reporting period determines which wagers, payouts, fills, credits, jackpots, and settlements belong in the calculation.
Federal minimum-control definitions provide a useful general statement: “actual hold percentage” is win divided by drop or coin-in. Nevada table-game controls likewise require statistical analysis showing drop, win, and win-to-drop hold percentage by table and game type. Those definitions show why a percentage without its base can be misleading.
| Area | Common activity base | What the percentage mainly describes |
|---|---|---|
| Slots | Coin-in | Win as a share of repeated wagering volume |
| Table games | Drop | Win as a share of value entering the table drop system |
| Sports wagering | Handle or settled wagers, depending on report | Win as a share of accepted or settled betting volume |
Those three ratios are useful inside their own systems. They should not be compared casually across products because the denominators behave differently.
Slot hold is close to a wagering ratio
For slots, coin-in records wagered credits, including credits won and then wagered again. Suppose a machine bank records $1,200,000 of coin-in and $84,000 of statistical win during the period:
\text{Actual slot hold}=\frac{84{,}000}{1{,}200{,}000}\times100=7.00\%This does not mean players inserted $1.2 million in fresh cash. A player can arrive with $100, win some credits, continue playing, and generate several hundred dollars of coin-in from that same bankroll. Coin-in therefore measures action more closely than cash inserted.
Slot operations also distinguish theoretical hold from actual hold. If the weighted theoretical hold for that same $1.2 million of coin-in is 8.2%, expected statistical win would be:
1{,}200{,}000\times0.082=98{,}400Actual win of $84,000 is $14,400 below that expectation, and actual hold is 1.2 percentage points below the theoretical rate. That does not by itself show that the games are configured incorrectly. Random variation, jackpots, a changing game mix, meter or reporting problems, and unusual promotional activity can all affect a period result. The proper operational response is to investigate persistent or material variances rather than infer a defect from one short sample.
When several slot groups have different theoretical holds, the correct floor expectation is weighted by coin-in. Suppose three groups produce $500,000 at 7%, $300,000 at 9%, and $200,000 at 12% theoretical hold:
H_w=\frac{500{,}000(0.07)+300{,}000(0.09)+200{,}000(0.12)}{1{,}000{,}000}=8.60\%Here H_w is the weighted theoretical hold. A simple average of 7%, 9%, and 12% would give 9.33%, but that would wrongly give equal importance to game groups that received very different volumes of play.
A 20% table hold does not mean a 20% house edge
Table games create the most common misunderstanding because table hold is often calculated against drop, not every wager placed. Suppose a blackjack pit records $100,000 of drop and finishes with $20,000 of win:
\text{Table hold}=\frac{20{,}000}{100{,}000}\times100=20\%Nothing in that calculation says blackjack had a 20% mathematical edge. The same chips can circulate through many hands after the original buy-in. A player who buys $500 in chips may make $5,000 or more in cumulative wagers before leaving. The casino's house edge applies to wagered action under stated rules and strategy assumptions; table hold applies to drop under the property's accounting definition.
Imagine the pit's estimated total action was $400,000. The $20,000 actual win would equal 5% of estimated action. Even that 5% is not automatically the game's true edge. Actual results contain variance, and blackjack results are also affected by the mix of player decisions, rule sets, side bets, and bet sizes. Table hold answers an operational question: how much win did the pit retain relative to money entering the table system?
This is why comparing table hold with slot hold can produce nonsense. A slot denominator repeatedly counts wagered credits. A table-drop denominator generally does not count every reuse of the same chips. A 20% table hold can coexist with game edges far below 20%.
Hold can rise while dollar win falls
Percentage and revenue are not interchangeable. Consider two slot banks:
| Bank | Coin-in | Actual hold | Gaming win |
|---|---|---|---|
| A | $100,000 | 10% | $10,000 |
| B | $500,000 | 7% | $35,000 |
Bank A has the higher percentage, but Bank B produces three and a half times as much win because it handled much more wagering volume. The same pattern appears in pits: a busy period can produce strong dollar win at a lower hold percentage, while a quiet period can show an unusually high hold percentage on small drop.
That is why managers normally read hold together with activity measures such as coin-in, drop, estimated decisions, average wager, occupied table hours, or handle. A percentage isolated from volume can make a weak business period look strong or a strong business period look weak.
Payout percentage is only the mirror when the base matches
For a slot game, theoretical hold and theoretical return to player use the same wager base, so the relationship is straightforward:
\text{RTP}=100\%-\text{theoretical hold}An 8% theoretical slot hold corresponds to 92% theoretical RTP. The same complement can be applied to actual slot hold if the numerator and denominator are defined consistently for the period.
But it is wrong to take a 20% blackjack table hold and announce that “players got back 80% of their wagers.” The denominator was drop, not total blackjack action. The arithmetic 100% − 20% is easy; the interpretation is false because the bases do not match.
Actual, expected, and realized hold answer different questions
Expected hold is the modeled result for a defined game mix and amount of activity. Realized hold or actual hold is what the property recorded during a real period. The two should converge only under conditions where the model, denominator, game mix, and sample are comparable.
A difference between actual and expected hold can come from ordinary variance. It can also expose something worth checking: an incorrect paytable, a meter issue, a coding problem, an unusual jackpot, inaccurate table ratings, promotional accounting, or a changed mix of play. The percentage is therefore both a performance measure and, in some environments, a diagnostic signal. It is not proof by itself of what caused the difference.
For a regulatory reference, 25 CFR 542.2 defines actual hold percentage using win divided by drop or coin-in and permits the measure to be calculated at different levels and periods. Nevada's current table-game internal-control standards similarly require statistical reporting of drop, win, and win-to-drop hold percentage by table and type of game; its slot procedures calculate actual slot hold from statistical win divided by coin-in.
Read a hold report with five checks
- Name the denominator. Coin-in, drop, handle, and settled wagers are not interchangeable.
- Define the numerator. Determine whether the report uses statistical win, accrual win, taxable win, or another adjusted figure.
- Match the period. A numerator containing settled results should not be paired casually with a denominator containing large amounts of unsettled activity.
- Check the sample. Short periods can be dominated by normal variance, especially for volatile products.
- Check the mix. A blended hold can move because customers shifted toward games with different theoretical rates even when no individual game changed.
Negative hold is possible during a period when players collectively win more than the casino against the chosen denominator. An unusually high positive hold is also possible in the short run. Neither result rewrites the underlying game mathematics.
The practical definition is therefore precise: hold percentage is gaming win expressed as a percentage of a specifically defined activity base. The percentage becomes informative only after you know what went into the numerator, what went into the denominator, and what period the calculation covers.