Video poker, the game that prints its price on the glass

A round is five-card draw against a pay table: you are dealt five cards from a standard 52-card deck, choose which to hold, and draw replacements once. There is no opponent and no bluffing; the machine pays fixed amounts for poker hands, from a pair of jacks up to a royal flush, and everything below the minimum pair loses.

That structure makes the game fully computable. Every deal has exactly 32 ways to hold, each with an exact expected value over the remaining 47 cards, and the published return of a pay table is the average of always choosing the best one. These are combinatorial results computed over the whole deal space; we state them rather than derive them, but unlike most casino figures they can be independently recomputed from the rules alone.

Variants are named by two cells of the pay table: what a full house and a flush pay per coin. The full-pay 9/6 Jacks or Better schedule returns 99.54% under optimal play, one of the lowest house edges on any casino floor. The 8/5 version of the same game, identical in every other respect, returns about 97.3%.

That gap is worth deriving in cost terms. A machine returning 99.54% keeps 46 of every 10,000 staked; one returning 97.3% keeps 270, nearly six times as much, for the same game at the same pace. Casinos routinely place both schedules side by side, and the pay table on the glass is the only thing that tells them apart. No other game moves its price this much while looking exactly the same.

The published return assumes you always make the mathematically best hold, and some of those holds are unintuitive: breaking a paying pair to chase a four-card flush is sometimes right, keeping a lone high card over a low pair never is. Full strategy charts for Jacks or Better run to several dozen ranked lines, and every deviation hands back part of the return.

Casual play typically gives up an extra fraction of a percent to a few percent depending on how far it strays. The honest framing: video poker is the casino game where skill matters most and where skill still cannot make the number positive. Optimal play on a 9/6 machine loses at 0.46%; every mistake only raises that.

Nearly two percentage points of the 99.54% return sit in a single hand: the royal flush, which arrives about once in 40,000 hands under optimal play and usually pays 800 to 1 at maximum coins. Between royals, the machine effectively plays two points tighter than its published figure.

This is the same volatility structure as a progressive jackpot slot in miniature: a low-edge average built from a long grind punctuated by one rare payout. It is also why max-coin play matters on these machines; most pay tables only unlock the 800 to 1 royal at full coins, and playing short coins quietly lowers the published return. A session bankroll sized for the average will not survive the variance; the session cost tool models the grind, not the royal.