Games · Video Poker

Video Poker: the machine that grades your play

The best-paying machine in the casino — if you play it right, and the machine tells you whether you did. Jacks or Better 9/6 returns 99.54% with perfect play, and full-pay Deuces Wild actually returns over 100% — numbers this site doesn't quote from anywhere: our engine enumerates every possible deal and every possible hold to derive them. Deal a hand, pick your holds, and the same solver grades your decision against the mathematically best one.

Credits $200.00
Your hand

Choose a paytable and deal.

Bet $1.25 (always 5 coins — the royal only pays its jackpot rate at max bet)

Free game, fun credits only — no real money. Verdicts come from the exact solver: every possible hold of your dealt hand is priced and yours is compared with the best one.

How this simulation works
Rules modeled
Standard draw video poker, single 52-card deck per hand. Five cards dealt; hold any subset; discards are replaced from the remaining 47 cards; the final five-card hand pays per the machine's schedule. Twelve schedules are modeled across three families: the Jacks or Better ladder from full-pay 9/6 down to the 6/5 found on cruise ships and airport floors; Bonus Poker 8/5, 7/5 and 6/5 (quad payouts split 80/40/25 by rank on every rung); and Deuces Wild full-pay, 'Not So Ugly', and Illinois/Airport (all four 2s wild, nothing below three of a kind pays). All pricing is at max bet (5 coins), where the natural royal pays 800 per coin instead of 250. Stakes are fun credits in integer cents.
Assumptions
Hands are independent (a freshly shuffled deck each deal). 'Perfect play' means the EV-maximizing hold among all 32 subsets, priced exactly over every possible draw. Max-bet basis throughout — betting fewer than 5 coins strictly lowers the return via the royal. Paytables vary by casino floor; these twelve are the standard published schedules, from each family's full-pay benchmark down the short-pay ladder casinos actually stock.
Mathematical basis
Every figure is exact, not simulated. The engine evaluates all C(52,5) = 2,598,960 possible deals, prices each of the 32 holds per deal by exact enumeration of draw completions (via a subset-aggregation + inclusion–exclusion core that a cross-check suite holds equal to direct enumeration on every hold to nine decimals), and keeps the best. The twelve resulting returns — from full-pay Deuces Wild's 100.7620% player edge down to 6/5 Jacks or Better's 94.9961% — each reproduced the published figure to six decimals and are committed as a versioned artifact the pages read. The Play tab's strategy verdicts and the Lab's simulation run the same solver; the Lab's Monte-Carlo only validates the exact figure, never redefines it.
Engine version
video-poker engine 1.0
Validation
Unit tests — both hand evaluators (standard + Deuces Wild) against oracle cross-checks, plus an exhaustive sweep of all 4,560 three-plus-wild hands pinning the full category distribution; all twelve paytables against the published schedules including royal jackpot tiers and Bonus Poker quad tiers; the reference solver against published strategy anchors; the fast core against the reference solver on every hold of sampled hands across all three paytables; and the committed RTP artifact pinned against the live paytables from both ends. 54 engine tests in CI plus the offline full-enumeration generator.
Last reviewed
Independent adversarial review (D-008) passed 2026-07-27: a separate reviewer's from-scratch implementation — own evaluators, own enumeration pipeline — matched the evaluator on all 2,598,960 hands in both modes with zero mismatches, matched hold EVs to 1e-10, and reproduced all three exact returns to ~1e-9 against Wizard of Odds' published figures. No critical or material findings; minor findings fixed with regression tests.