actfap

Video Poker Return Calculator

Type in any pay table and get the exact optimal-play return, solved over all 2,598,960 deals.

This compares pay tables at perfect play, on four non-wild games. It does not tell you how you personally will do on one, it is not a strategy trainer, and it does not cover Deuces Wild, which has its own page.

The plainest schedule in the casino and the one every other game is a variation on. Only two numbers move between machines: the full house and the flush.

Pay table on the glass, per coin

Full pay. Increasingly rare outside Las Vegas locals casinos.

9/5 and 8/6 differ in 2 numbers on the glass and return within 0.06 points of each other, 98.45% against 98.39%. Two schedules that look different can pay almost the same, which matters as much for choosing a seat as two that look the same paying differently.

Exact optimal-play solution for 9/6 Jacks or Better, computed here over every one of the 2,598,960 five-card deals.

What a hand actually returns

55%

of hands return nothing

Typical result: nothing on the middle hand

Every outcome, and where the return comes from
HandChanceShare of return
Royal flush800 for 10.0025%1 in 40,3912.0%
Straight flush50 for 10.011%1 in 9,1480.5%
Four of a kind25 for 10.24%1 in 4235.9%
Full house9 for 11.15%10.4%
Flush6 for 11.10%6.6%
Straight4 for 11.12%4.5%
Three of a kind3 for 17.44%22.4%
Two pair2 for 112.9%26.0%
Jacks or betterstake back21.5%21.6%
Nothingnothing54.5%0.0%
Return at optimal play99.54%

The number on the glass, and it is an average over every hand. It assumes the best hold on all 2,598,960 deals, which no player finds every time. Note that the chances in the table above are not fixed furniture: a lower flush makes some four-card-flush holds the wrong play, so a worse machine changes how often each hand arrives as well as what it pays.

Exact enumeration, not a simulation and not an interpolation. Every deal is evaluated against all 32 holds and every possible draw, and the best hold is taken. See how this is calculated, below.

House edge
0.46%
Royal flush
1 in 40,391
Return without a royal
97.57%

How much rides on picking correctly

Second-best hold on every deal

4.7c

per coin staked, given up

That one policy, in money

$142

an hour, at 25c a coin and 600 hands an hour

Every deal has one best hold and one second-best, and this engine already finds both while it searches. Averaged over all 2,598,960 deals, taking the second-best hold instead of the best gives up 4.7c per coin staked. This is not an estimate of how often a player actually errs, and no real player takes the worst-of-two-best choice on every hand. It is the cost of that one specific policy, not a ceiling on every kind of misread (holding outside the top two costs more): at 25c a coin and 600 hands an hour, always taking the second-best hold would cost an extra $142 an hour beyond the table's own edge, which is the size of the gap this particular wrong read opens on this pay table.

What this table costs to sit at

5 coins a hand, which is $1.25 at this denomination. Max coin is assumed throughout: below it the royal pays 250 rather than 800 and every return on this page is too high.

Expected cost / hour

$3.42

Bet per hand$1.25Hands / hour600Through the machine / hr$750

Over 3 hours

$10.26

This is the best Jacks or Better schedule on the list (9/6), so there is no better version of this game to walk to. It still has a cost, and it is the figure above.

Every figure here is an expected cost, not a forecast of your night, and it assumes perfect play on every hand. The section above shows what always taking the second-best hold costs, not a ceiling on every kind of wrong read (holding outside the top two costs more), and not an estimate of how often real players actually misread a hand. Video poker is gambling: play only with money you can afford to lose, and only where it is legal for you to do so. 21+.

Written and reviewed by Alexander Korsager · last reviewed October 7, 2026 · corrections

The idea

What this tool shows

Two machines side by side can look identical and return differently, by four and a half points on Jacks or Better alone. The whole difference is usually two numbers: what a full house pays and what a flush pays, per coin, below max bet. The ladder above prices every schedule this page solves, from full pay down to the worst in common circulation, so the gap is a curve you can read rather than a few figures to hold in your head, and it also carries the less obvious half: two schedules that look different on the glass can return almost the same.

  1. This calculator does not look those figures up.Type in whatever the machine in front of you pays, on any of four games, and it solves the table from scratch: every one of the 2,598,960 five-card deals, evaluated against all 32 ways to hold, against every possible draw. That matters because optimal strategy is a function of the pay table. Lower the flush and some four-card-flush holds stop being the right play, which changes how often every hand arrives, not only what it pays. A calculator that applies one published strategy to every schedule cannot see that, and a table nobody has published a strategy for cannot be answered at all.

  2. The second number here is the one a player at a bank of machines is actually choosing between.A percentage is not a decision. What the 8/5 machine costs you an hour more than the 9/6 one, at the denomination you are about to feed it and the speed you play, is a decision, and this page prints it in dollars next to the return.

FAQ

Questions players ask

What is a 9/6 machine?

One where a full house pays 9 for 1 and a flush pays 6 for 1 at less than max coin. It is the 'full pay' Jacks or Better schedule and returns 99.54% with optimal play. Every other schedule this game ships, from 9/5 down to 6/5, returns less than that; the ladder above prices the whole set rather than a few named ones, and the calculator will solve any table you type in too.

Where does the return figure come from?

It is computed on this page, not transcribed. For every possible five-card deal the calculator works out the expected value of all 32 hold patterns against every possible draw, takes the best one, and averages over all 2,598,960 deals. That is what 'optimal play' means, and it is why an edited pay table gets a real answer rather than an interpolated one.

Does this account for the max-coin royal bonus?

Yes, the royal flush defaults to 800 for 1, which is the max-coin rate. Always bet max coins on video poker: at less than max the royal only pays 250 for 1, and every other row stays the same, so the return drops. Change the royal row above to 250 on any schedule and the panel resolves that schedule's return at the lower rate, the same enumeration as the headline return rather than a lookup. Subtract it from that schedule's row in the ladder above to read off the exact drop, which shifts a little from one schedule and game to the next.

Why is Double Bonus shown returning more than 100%?

Because full-pay 10/7 Double Bonus genuinely does, at 100.17%, under flawless play at max coin. That is a real schedule and it is also close to extinct, and the gap between perfect play and good play on Double Bonus is wider than on Jacks or Better, so the theoretical edge is not something a person collects by sitting down. Read it as what the machine gives up, not as income.

Is Deuces Wild covered?

Not on this page. The wild card changes how hands are classified rather than just what they pay, so it needed a separate evaluator rather than a fifth game bolted onto this one. It has its own page now, Deuces Wild Video Poker, sharing this tool's solver but with its own classifier and its own four schedules. This page stays scoped to Jacks or Better, Bonus Poker, Double Bonus and Double Double Bonus.

Methodology

How this is calculated

The return is solved by exhaustive enumeration, not by simulation and not by applying a published strategy.

Every one of the C(52,5) = 2,598,960 deals is evaluated against all 32 hold patterns; each hold's expected value is its total payout over every legal draw divided by the number of draws; the best hold is taken; and the results are averaged.

The hand frequencies shown are the ones those chosen holds actually produce for the pay table on screen, so they move when the pay table moves. The hourly cost is that return applied to a stake and a speed: hands per hour, times five coins at the chosen denomination, times one minus the return.

Common misreadings4
  • A 99.54% return is not 99.54% of hands giving something back. Over half of hands return nothing, and about a fifth return exactly the coin staked, which is not a win. Under a quarter of hands come out ahead, on full-pay Jacks or Better, the best schedule that game has.
  • The advertised return includes the royal flush, which comes up about once in 40,000 hands. A player who never hits one is playing a game that returns closer to 97.6%, and the page shows that figure next to the headline for exactly this reason.
  • Choosing a better pay table does not mean winning more often. It changes what hands pay, and it shifts how often each hand arrives only because the best hold changes; it does not turn a losing game into a winning one.
  • A theoretical return above 100% is not an income. It is what the machine concedes to a player who never makes a mistake, before the cost of the hours it takes to collect it, and any real error rate puts it back under water.
Assumptions6
  • Five cards, one draw, played at the best available hold on every deal. The solver never guesses: it evaluates all 32 holds on each deal and takes the maximum.
  • The schedules on the list are solved ahead of time and appear immediately. A table you have typed is solved in your browser, which takes a second or two, and nothing on the page is shown for it until that finishes. No figure here stands in for a solution that does not exist yet: where there is nothing to show, the page says so instead of printing a number.
  • A pay row left blank is solved as a hand that pays nothing, and the page says which rows those are while any are blank. That is a real schedule and it gets a real answer, but it is not the machine a reader half way through retyping a payout is looking at, which is why it is labeled rather than silently priced.
  • Pays are 'for 1' per coin, and the stake is five coins a hand. The royal flush defaults to 800, its max-coin rate, so every figure describes a max-coin player and overstates the return for anyone betting fewer coins.
  • The default of 600 hands an hour is the same pace this site's house-edge comparison uses for video poker. It is an input, not a constant, and the hourly cost scales linearly with it.
  • Cost figures are expected values over the long run. A single session is dominated by whether a rare hand lands, and the distribution above is there to show how rare.
Limits4
  • Perfect play is the ceiling, not a forecast. The mistake-cost figure on the page is the cost of one specific policy, always taking the second-best hold, not a ceiling on every kind of misread: a hold outside the top two costs more. It is computed exactly from the same enumeration as the return, it is not a model of how often real players actually misread a hand, and it should not be read as an estimate of typical player loss beyond the return figure itself.
  • The distribution describes one hand. It does not model a session, a bankroll, or the chance of surviving long enough to reach a rare hand.
  • Only the four non-wild games listed are modeled. Deuces Wild classifies hands differently (its own page solves it separately) and other wild-card variants further still, so none of these numbers carry across to them.
  • Progressive royal jackpots are not modeled. Where a royal pays more than the number in the table, the true return is higher than what is shown, and the royal's share of it larger still.
Evidence

The solver is checked against figures computed independently of this site: the optimal-play returns for 9/6, 9/5, 8/6, 8/5, 7/5 and 6/5 Jacks or Better, 8/5 Bonus Poker, 10/7 Double Bonus and 9/6 Double Double Bonus, all matched to six decimal places, on four games with different pay structures.

The hand-classification table is checked against the textbook five-card frequencies anyone can verify without this repo (4 royal flushes, 36 straight flushes, 624 four of a kinds, 3744 full houses, 5108 flushes, 10200 straights).

The optimal-play hand frequencies for 9/6 Jacks or Better reproduce the published table to all eight decimal places it is published to, having been computed rather than copied, which is a second check on the solver: a right return can in principle sit on a wrong distribution, and this says it does not.

A passing test means the code matches the model, not that the model matches a specific machine.

What the figures do while there is no solution is pinned separately, in panel.test.ts, and it is pinned because until 2026-09-10 it was wrong: the first edit away from a preset left the page with nothing solved, and every figure was computed from a return of zero rather than withheld, so for the second or two the solve took it printed a house edge of 100.00%, an hourly cost equal to the whole turnover, and a comparison reading 'an extra $747 an hour'.

The same commit made the ladder read its payouts from the vector the solver was handed rather than from the raw fields, so a blank row shows the nothing it is being solved as instead of NaN.

Gambling can cause serious harm. The National Problem Gambling Helpline is free, confidential and open 24/7, call or text 1-800-MY-RESET (1-800-697-3738), or chat at ncpgambling.org/chat.

Read up

Go deeper

Related tools

Try next

We’d like to use privacy-friendly analytics (Google Analytics, Vercel Analytics and Speed Insights) to see which tools and pages are actually useful, so we can improve them. Nothing is sold, and nothing loads unless you say yes. See our privacy policy for details.