Video poker hold percentage is easiest to understand when three different numbers are kept separate: theoretical hold, actual hold, and casino win dollars. They are related, but they answer different questions.
Theoretical hold describes the long-run margin built into a particular paytable under a stated strategy assumption. Actual hold describes what the machine or bank really retained over a measured period. Win dollars describe the money result behind that percentage. A manager who looks at only one of those numbers can misread an otherwise normal game.
The three numbers should never be collapsed into one
For a video poker game with a theoretical RTP of 98.5%, the complementary theoretical hold is:
[ \text{Theoretical Hold}=1-0.985=0.015=1.5% ]
If the game records $500,000 of coin-in, a simple theoretical-win estimate is:
[ $500{,}000\times 0.015=$7{,}500 ]
That $7,500 is an expectation, not a promise. The actual machine result can be far above or below it because video poker payouts are uneven and rare top hands contribute a meaningful part of the return.
Actual hold uses the result that really occurred:
[ \text{Actual Hold}=\frac{\text{Casino Win}}{\text{Coin-In}} ]
If the same machine generated $500,000 of coin-in and finished with $20,000 of casino win, its actual hold was 4%. If it finished at negative $5,000 after an unusually strong payout period, its actual hold was -1%. Neither outcome changes the paytable’s theoretical hold.
Coin-in is wagering volume, not cash inserted
A major source of confusion is the denominator. Coin-in is the amount wagered through the game, including recycled credits. It is not simply the amount of currency placed in the machine.
A player can insert $100 and create $1,000 or more of coin-in by repeatedly wagering credits that are won and replayed. For hold analysis, the $1,000 of wagering volume is the relevant denominator.
This is why casino win and player buy-in should not be compared directly when evaluating machine hold. A player may buy in for $200, cycle thousands of dollars in wagers, and cash out with $150. The casino won $50 from that individual session, but the game generated far more than $200 of coin-in.
Video Poker Accounting explains the transaction flow in more detail, while Video Poker Meter Readings covers the evidence used to reconcile meters and system reports.
Video poker hold is unusually sensitive to strategy
In a conventional reel slot, the player’s choices usually do not change the mathematical return of a completed spin. Video poker is different because the player chooses which cards to hold before the draw.
That means the advertised or calculated theoretical return is tied to a strategy assumption. A paytable that returns 99% under optimal play does not guarantee that every player will achieve a 99% long-run return. Repeatedly discarding strong made hands, chasing visually attractive but inferior draws, or using a strategy chart for the wrong paytable can lower the player’s realized expectation and raise the casino’s effective margin against that player.
The key distinction is:
- Paytable hold comes from the game’s mathematical design under the stated strategy model.
- Player-specific effective hold can be higher when decisions are systematically suboptimal.
- Actual hold is still the observed accounting result, which includes both strategy effects and ordinary variance.
The deeper paytable-and-strategy relationship is covered in Video Poker House Edge.
One royal flush can dominate a short reporting period
Video poker results are lumpy. A large share of total return can come from relatively rare hands, especially the royal flush on games with a strong top award. That creates a practical management problem: a short period can look dramatically different from theory even when the game is functioning exactly as designed.
Consider two identical machines with the same paytable and $100,000 coin-in each.
- Machine A has no major top-hand payout during the period and wins $4,000: actual hold = 4%.
- Machine B pays a large royal and loses $1,500 overall: actual hold = -1.5%.
The machines did not suddenly acquire different theoretical paytables. They experienced different samples.
This is why managers should not ask only, “What is the hold?” They should ask:
- How much coin-in produced the result?
- Over what time period?
- What is the theoretical hold for the active paytable?
- Were there major jackpots or hand-paid awards?
- Is the result isolated to one device, one bank, one denomination, or the whole category?
- Is the deviation persistent as volume grows?
A percentage without volume can be almost meaningless
Suppose one machine holds 8% on $2,000 coin-in while another holds 1.2% on $1,000,000 coin-in.
The first wins:
[ $2{,}000\times0.08=$160 ]
The second wins:
[ $1{,}000{,}000\times0.012=$12{,}000 ]
The 8% figure sounds much stronger, but the second game generates far more casino win. For floor decisions, percentage and volume have to be read together.
The same logic applies when comparing denominations. A quarter game, a dollar game, and a multi-denomination cabinet can produce very different coin-in, player profiles, and volatility even if their theoretical holds are similar.
Theoretical hold belongs to the active configuration
Casinos should not treat “Jacks or Better,” “Deuces Wild,” or any other family name as a sufficient hold description. Paytables can differ materially inside the same game family.
A bank that changes from one approved paytable to another changes the game’s theoretical economics. The relevant comparison is therefore not simply this month versus last month. Management should confirm which paytable was active, when it changed, and whether the reporting system associates the correct theoretical percentage with that configuration.
Nevada’s current slot minimum internal controls explicitly require slot-machine computer data to include theoretical hold percentage alongside coin-in, payouts, and win amounts. They also require controls around changes to theoretical hold data. The Nevada Gaming Control Board slot MICS are a useful primary reference for how theoretical and actual device information is kept distinct in casino records.
Actual RTP and actual hold are complements when measured consistently
If the same turnover and payout scope is used, actual RTP and actual hold are two sides of the same result:
[ \text{Actual RTP}=\frac{\text{Player Winnings}}{\text{Turnover}} ]
[ \text{Actual Hold}=1-\text{Actual RTP} ]
For example, if $1,200,000 of turnover produces $1,085,000 returned as player winnings, actual RTP is approximately 90.42% and the corresponding actual hold is approximately 9.58%.
The UK Gambling Commission uses exactly this kind of turnover-and-winnings calculation when describing live RTP monitoring. Its RTP calculation guidance also emphasizes that actual performance can differ from designed RTP over a finite period.
The scope must match. If one report treats certain jackpots, promotional awards, or linked-progressive contributions differently from another, subtracting one percentage from 100% may not create a valid apples-to-apples comparison.
Handpay method is not the cause of a hold swing
A common reporting mistake is to blame “handpays” when hold falls. The payment method is not the mathematical event. The underlying award is.
A jackpot may require attendant verification and manual payment because of its size or the machine’s configuration. That can make the event operationally visible, but the payout belongs in the game’s result regardless of whether credits were paid automatically, by ticket, or by an attendant.
The correct management question is not “Did we have many handpays?” It is “What awards occurred, were they properly recorded, and do the meters and system reports reconcile?”
That is an accounting and controls question, separate from whether the paytable is performing normally.
Strategy quality can make actual hold persistently higher than paytable hold
Variance explains short-term movement, but player behavior can create a more persistent gap. If a player population makes repeated strategy errors, the casino may realize a higher long-run hold than the optimal-strategy paytable figure suggests.
This does not mean the game changed. It means the actual decisions were different from the decisions assumed in the theoretical-return calculation.
For that reason, a useful casino analysis can separate:
- theoretical hold for the configured paytable;
- actual hold for the device or bank;
- coin-in volume;
- large-award history;
- denomination and game mix;
- player strategy quality where it can be estimated responsibly;
- time since the last paytable or configuration change.
Managers should be cautious about reverse-engineering individual skill from aggregate hold. A bank can overhold because of ordinary variance, player mistakes, or both.
A simple variance-to-theory dashboard
A practical management view can start with four calculations:
[ \text{Theoretical Win}=\text{Coin-In}\times\text{Theoretical Hold} ]
[ \text{Actual Hold}=\frac{\text{Actual Win}}{\text{Coin-In}} ]
[ \text{Hold Variance}=\text{Actual Hold}-\text{Theoretical Hold} ]
[ \text{Win Variance}=\text{Actual Win}-\text{Theoretical Win} ]
Suppose a bank records $2,000,000 coin-in, 1.5% theoretical hold, and $44,000 actual win.
- Theoretical win = $30,000.
- Actual hold = 2.2%.
- Hold variance = +0.7 percentage points.
- Win variance = +$14,000.
Those numbers show that the bank overheld during the measured period. They do not explain why. The next step is to examine volume, large awards, game mix, paytable changes, and data integrity rather than jumping straight to “players were bad” or “the machines were tight.”
Hold is a diagnostic, not a verdict
Video poker hold percentage is useful because it connects wagering volume to casino result. It becomes misleading when it is treated as a fixed property of a short period or as proof that a machine is behaving fairly or unfairly.
Theoretical hold belongs to the paytable and strategy assumption. Actual hold belongs to a defined sample of real play. Win dollars show the economic scale. Variance explains why the observed result can wander. Player decisions explain why a video poker population may not realize the optimal-strategy return.
Read all of those together. A hold number by itself is only the beginning of the analysis.