Games · Double Down Madness
Double Down Madness: one card, unlimited doubles, and a 22 that isn't a bust
You're dealt one card. From there you can hit, stand, or double — and double again, and again, and still hit afterwards. It is the most permissive doubling rule on any casino floor, and it is not a gift: the house pays for it with a single line of small print — a dealer 22 pushes instead of busting. That one rule turns roughly 7% of your wins into refunds. Played well on the best paytable it costs 0.95%; on the wrong paytable it costs more than double.
Press Deal. You'll get one card — then hit, stand, or double as often as you like.
The dashed gold line is the published house edge for the selected version. Watch the average final wager settle near 1.57× — that's the unlimited doubling at work, and it's why the two edge figures differ so much.
Our strategy is derived by the engine, not copied from a published chart. Measured across six million hands it lands on the published Version 1 figure within noise, and within 0.08 points on Versions 2–3 (the residual runs in the safe direction — it can only overstate the house edge). Expect the curve to wander: the geometrically compounding wager makes per-hand variance unusually large, so even a million hands can sit a tenth of a point either side.
The trade at the heart of the game
You get something genuinely rare — unlimited doubling, and you may still hit after you double. The house takes it back with one line of small print: a dealer 22 is a push, not a bust. Everything below is that trade, priced.
The three licensed versions
Casinos choose a version; they differ only in the blackjack payout, and it matters a lot. Two figures are quoted because unlimited doubling makes them diverge: the house edge is per your opening wager, the element of risk is per dollar actually put at risk.
| Version | Blackjack pays | House edge | Element of risk |
|---|---|---|---|
| Version 1 ⭐ | Suited BJ 2:1 · unsuited 3:2 | 0.95% | 0.61% |
| Version 2 | Suited BJ 3:2 · unsuited 3:2 | 2.07% | 1.32% |
| Version 3 | Suited BJ 3:1 · unsuited 1:1 | 2.07% | 1.32% |
Version 1 costs less than half what Versions 2 and 3 do. Before you sit down, look at the felt: if a suited blackjack doesn't pay 2:1, you're playing the expensive game.
Where the dealer's 22 comes from
The push-22 rule bites hardest against the upcards you were happiest to see. A dealer 6 busts often — but a healthy slice of those "busts" are 22s that now push instead of paying you.
Read as: the share of hands where the dealer lands on exactly 22. Those are the wins the rule quietly converts into pushes.
The Push 22 side bet
If the rule is going to happen anyway, the felt invites you to bet on it. Don't.
At 11:1 you need it to land more than 8.33% of the time to break even; it lands 7.36% of the time. That gap is a 11.72% house edge — roughly 12× the cost of the main game. It's the most expensive thing on the table.
Published figures: Wizard of Odds (six decks). Our engine's independently computed dealer-22 probability is 0.07357, which reproduces the published side-bet edge to within a rounding step — see methodology.
How this simulation works
- Rules modeled
- Double Down Madness (Light & Wonder; debuted July 2024) on six ordinary 52-card decks. The player is dealt ONE card and may Stand, Hit, or Double on every card received; doubling is unlimited, the player may still hit after doubling, and re-doubling is GEOMETRIC — each double doubles the total wager (1× → 2× → 4× → 8×, per the published rule that each additional wager is double the previous one). Doubling on a lone ace buys exactly one card and ends the hand; hitting a lone ace continues normally. No splitting and no surrender. The dealer takes two cards, peeks for blackjack, and hits all 17s including soft 17. A dealer total of exactly 22 PUSHES all standing wagers; a player blackjack still beats a dealer 22. Blackjack is ace + ten-value in the first two cards, paid by version: V1 suited 2:1 / unsuited 3:2, V2 both 3:2, V3 suited 3:1 / unsuited 1:1. Optional Push 22 side bet pays 11:1. Stakes are fun credits.
- Assumptions
- Hands are independent within a penetrated shoe (reshuffle at 75% dealt). The strategy is DERIVED by the engine's own EV recursion, not transcribed from a published chart. Because re-doubling is geometric, the hand's value scales exactly with the wager, which makes the recursion clean: double beats hit precisely when the drawn-to position has positive per-unit value. The recursion is an infinite-deck model (composition-independent) and averages the suited/unsuited blackjack odds in decisions (settlement pays suitedness exactly), which leaves a small residual on Versions 2–3 — about 0.06–0.08 percentage points, in the safe direction (it can only overstate the house edge). It also independently reproduces the published strategy's signature lines: always double a lone ace or ten in Version 1, but HIT ace-against-ace in Versions 2 and 3. Note that per-hand variance is unusually large here because the wager compounds geometrically — a single million-hand run can sit a tenth of a point either side of the mean. Fun credits only, no real money.
- Mathematical basis
- Two verification layers against published figures (Wizard of Odds, six decks). The Push 22 side bet is computed from the engine's own dealer-outcome distribution and matches almost exactly: our P(dealer makes exactly 22) = 0.07357 against the 0.07353 implied by the published 11.76% edge at 11:1 — four decimals, independently confirming the dealer model. The main game is verified by seeded Monte-Carlo, measured properly across 6 seeds × 1,000,000 hands per version: Version 1 observed 0.986% (standard error 0.038 points) against the published 0.95% — statistically indistinguishable — with the element of risk at 0.626% vs 0.61% and the average final wager at 1.57 exactly. Versions 2 and 3 observe 2.15%/2.13% against 2.07%, a ≤0.08-point residual from the derived strategy's approximations, in the conservative direction. Worth recording: an earlier build misread the doubling rule as adding one unit per double and measured ~0.19 points high — the published average-wager figure (1.57) is what exposed the misread. Note too that this game quotes two very different edge conventions because of the compounding wager: house edge is per opening wager, element of risk is per dollar actually risked.
- Engine version
- double-down-madness engine 1.0
- Validation
- 22 unit tests + a 3×2M-hand slow verification — ordinary 52-card shoe composition, the dealer distribution as a proper distribution across every upcard with 22 separated from a true bust, the published Push 22 probability and side-bet edge, every settlement rule (dealer 22 pushes a standing wager but a busted player still loses; a player blackjack beats a dealer 22; each version's suited/unsuited blackjack odds; blackjack paid on the full committed wager; a three-card 21 is NOT a natural), the ace-start one-card rule, published strategy checkpoints, determinism, and MC convergence to all three published edges.
- Last reviewed
- Verified against every published anchor the game offers — the side-bet probability, three separate paytable edges, both edge conventions, and the average final wager. Two real modelling bugs were found and fixed by probing against those anchors rather than by inspection: a single-boolean softness collapse that mis-scored soft hands, and a solver blind to blackjack bonuses that systematically under-doubled. Built on the QC'd blackjack engine's primitives. Last reviewed 2026-07-22.