A video poker random number generator does not decide whether a player deserves to win. Its job is narrower: produce random values that the approved game software can use to select cards. The player then chooses which cards to hold, replacement cards are selected under the game rules, and the final hand is paid according to the displayed paytable.
That process can be technically sound while the player still loses quickly. Game integrity means the machine follows its approved rules and records the transaction correctly. It does not mean results will look balanced, that every paytable is generous, or that a short session will resemble the theoretical return.
The RNG is therefore only one part of the integrity question. The complete chain includes random selection, card mapping, the hold-and-draw sequence, paytable configuration, software authentication, access controls, meters, event logs, and game recall.
The RNG is not a deck of cards
An RNG produces numbers. The game program converts those numbers into card selections according to its approved design.
A simplified video poker round looks like this:
- The player commits a wager and starts the deal.
- Random values are used to select five distinct cards for the initial hand.
- The machine displays those cards and waits for the player’s decision.
- The player marks zero to five cards as held.
- After the player starts the draw, the game selects replacement cards for the discarded positions.
- The final five-card hand is compared with the active paytable.
- Credits, meters, and game-history records are updated.
The exact software implementation can vary by manufacturer and jurisdiction. Some systems may construct a shuffled virtual deck; others may repeatedly map random values to unused cards. What matters is that the approved process produces the required card probabilities, does not duplicate a card within the same hand, and cannot be influenced by information that should be irrelevant to the outcome.
For a broader tour of the cabinet, meters, credits, and TITO functions, see how video poker machines work. This page stays with the integrity of the deal and draw.
Why the hold decision creates a special integrity requirement
Video poker is not a passive reel game. The initial hand creates a decision point, and the player’s choice changes the set of possible final hands.
Suppose the deal is:
A♠ K♠ Q♠ 7♦ 2♣
The player may hold the three high spades and draw two cards, hold only the ace, or make another legal choice. Each decision leaves a different set of possible outcomes. The correct choice depends on the exact variant and paytable, as explained in hold-or-draw decisions.
Integrity requires the machine to respond to the hold decision under the approved rules. It must not secretly choose a losing replacement hand first and then make the displayed draw fit that result. It also must not reward or punish a particular hold because of the player’s identity, recent wins, loyalty status, or button rhythm.
This distinction matters because players often confuse three separate events:
- The initial deal, which is random.
- The hold decision, which is chosen by the player.
- The replacement draw, which is random but constrained by the cards already in the hand and the cards held.
A correct decision can lose. A poor decision can hit a royal flush. The outcome of one hand does not prove whether the decision or the random process was sound.
What a regulator can require of the random-selection process
Technical rules differ by jurisdiction, so one regulator’s exact numbers should not be presented as a universal worldwide standard. Nevada provides a useful concrete example.
Its current Technical Standard 1 requires the random-selection process to meet a chi-squared goodness-of-fit test at stated confidence limits. For software RNGs, it also requires a non-static initialization seed, cycling at a minimum average rate of 100 times per second, and no selection of RNG values for future play. Most directly for video poker, the standard says replacement cards must not be determined before the player selects the holds and initiates the draw.
The same standard requires protection against outside electrical or communication influences. It also says that if a game uses a predetermined set such as a shuffled deck, information about that set must not be accessible.
These are jurisdiction-specific controls, not a promise that every country uses identical wording. They do show why common player theories are weak:
- Waiting three seconds does not create a knowable card.
- A loyalty card is not supposed to switch the player to a worse deck.
- A machine is not permitted to store future draws for use when it needs to recover previous losses.
- A replacement card should not already be fixed before the player makes the hold decision.
The dedicated button-timing myth and player-card myth pages examine those claims from the player’s point of view.
Randomness and return are different controls
A fair RNG does not tell you whether the game is a good wager. The active paytable and the player’s strategy determine the theoretical return.
For a particular game and strategy, theoretical return can be written as:
RTP = Σ(pᵢ × rᵢ)
where:
- pᵢ is the probability of final result i under the chosen strategy;
- rᵢ is the amount returned per unit wagered for that result;
- Σ means the products are added across all possible final results.
If a game has a theoretical return of 99.5%, the corresponding mathematical house edge is:
House edge = 1 − 0.995 = 0.005 = 0.5%
On 2,000 units of total coin-in, the long-run expected loss would be:
2,000 × 0.005 = 10 units
That calculation is an average over repeated play under the assumed strategy. It does not predict a ten-unit loss in a particular session. A player may finish far ahead or far behind because video poker results have substantial variance.
The key point is that RNG integrity and mathematical value answer different questions:
| Question | What answers it? |
|---|---|
| Are cards selected through the approved random process? | RNG and game-program integrity |
| Which winning hands are paid and by how much? | The displayed and approved paytable |
| Is the player choosing the highest-value hold? | Strategy for that exact game and paytable |
| Why did this session swing so sharply? | Variance and the sequence of actual outcomes |
| What is the long-run casino advantage? | Paytable probabilities plus player strategy |
A perfectly random machine can have a weak video poker paytable. A strong paytable can be damaged by repeated strategy errors. Neither problem is evidence that the RNG is biased.
Integrity extends beyond the random numbers
A machine could produce statistically acceptable random values and still have an integrity problem elsewhere. A complete review may need to examine:
- whether the installed software matches the approved version;
- whether the correct paytable is active for the displayed game and denomination;
- whether the initial hand, holds, draw, final hand, wager, and award are recorded correctly;
- whether software changes and remote access are authenticated and logged;
- whether meters increase by the correct amounts;
- whether a malfunction or communication failure put the game into the required error state;
- whether game-recall information matches what appeared on the screen;
- whether the cabinet, logic area, and access points show signs of unauthorized entry.
This is why machine testing and certification is a separate subject. Certification is not one statistical test on a stream of numbers. A laboratory or regulator may review source code, random-output data, game rules, paytables, software identifiers, checksums, hardware behavior, accounting functions, and error handling.
Gaming Laboratories International’s current RNG submission guidance, for example, asks for production-like final-outcome collection, raw output when required, source-code documentation, hardware details, and file checksums. That is evidence of a reviewed implementation, not a guarantee that an individual player will receive a smooth sequence of wins and losses.
What official Nevada controls make visible after a dispute
The Nevada Gaming Control Board’s Technical Standard 1 illustrates controls that extend beyond the RNG. It requires game-play history capable of showing the outcome, intermediate steps such as the hold-and-draw sequence, credits available, bets placed, amounts paid, and cash-outs. It also contains software-authentication, change-log, and remote-access requirements for applicable systems.
Those records are more useful in a dispute than a player’s memory of how the hand “felt.” They can help answer concrete questions:
- What five cards were originally displayed?
- Which cards were marked as held?
- What replacement cards appeared?
- What was the denomination and number of credits wagered?
- Which paytable was active?
- How many credits were awarded?
- Did the machine enter a tilt or error state?
Game recall does not prove that an unlucky draw was statistically unusual enough to indicate a defect. It creates an auditable record of the transaction so staff and regulators can compare the recorded event with the approved rules.
A bad result is not the same as a malfunction
Consider a player dealt four cards to a royal flush who draws an unrelated card. The miss is disappointing, but it is the ordinary result most of the time. The machine does not owe a near-miss conversion because the player has already missed several royals that day.
A technical review becomes more appropriate when there is a specific observable inconsistency, such as:
- a held card disappearing or an unheld card remaining;
- a final hand being paid at an amount different from the displayed paytable;
- credits not being added after a documented win;
- duplicated cards appearing in a single standard 52-card hand;
- the screen, printed ticket, meter, and game recall showing conflicting amounts;
- an error message or lockup during the hand;
- the game title or paytable changing without a clear transition before the wager.
If that happens, stop play and note the machine number, time, denomination, wager, displayed cards, and error message. Ask for an attendant or supervisor before the screen is cleared. Procedures and dispute rights vary by jurisdiction, but precise facts are far more useful than saying only that the machine was “cold.”
What the player can verify without seeing the source code
A player cannot independently audit the RNG from a few hands. Short samples are dominated by normal variation, and the absence of a royal flush says almost nothing about the quality of the random process.
The player can still verify several practical details:
- Read the full paytable before wagering. The game name alone does not establish the return.
- Confirm denomination and credits bet. A correct hand can appear underpaid when the player misreads coin value or max-coin rules.
- Use strategy for the exact variant. The same cards can have different best holds under different paytables.
- Separate tracking from outcome generation. A player card records play; it should not alter the draw.
- Judge disputes by reproducible facts. A wrong award or inconsistent game recall is evidence. A losing streak by itself is not.
- Treat certification as process assurance, not profit assurance. Approved randomness can still produce severe losses.
The useful question is not, “Did this machine give me a fair-looking hour?” It is, “Did the approved software select and settle this hand according to the rules, and is there an auditable record if something appears wrong?”
That is what video poker game integrity is designed to answer.