- Rules modeled
- Standard Let It Ride, single 52-card deck, independent hands. The player makes three equal bets and receives three cards; two community cards complete every hand's final five. After seeing the three cards the player may pull back bet 1; after the first community card is revealed the player may pull back bet 2; bet 3 always rides. Each riding bet is paid by the final 5-card hand on the standard table, pair of tens or better 1:1, two pair 2:1, three of a kind 3:1, straight 5:1, flush 8:1, full house 11:1, four of a kind 50:1, straight flush 200:1, royal flush 1000:1, and loses on anything less. Stakes are fun credits in integer cents.
- Assumptions
- Hands are independent (a freshly shuffled deck each round). Optimal play is defined exactly: a bet rides when the exact expected value of a riding unit, averaged over every possible completion, is positive, and is pulled otherwise. The Play tab's hints and the Lab's strategy use those same exact EVs. The 5-Card Bonus side bet is omitted: its paytables vary wildly by casino (published edges ~14–37%), so there is no honest single number to show. Fun credits only, no real money.
- Mathematical basis
- Every figure is EXACT, Let It Ride has no dealer, so the entire game is enumerable. The house edge (3.5057% per base bet), element of risk (2.8648%), ride rates (bet 1 rides 7.3% of hands, bet 2 15.1%), and final-hand distribution are derived by backward induction over all 22,100 three-card starts × 49 × 48 completions (~52M states). The Lab's Monte-Carlo (one continuous seeded RNG stream) only validates the derivation, never defines it.
- Engine version
- let-it-ride engine 1.0
- Validation
- 16 unit tests + an opt-in slow test; the 5-card evaluator (all categories, the tens-or-better split, the wheel and steel wheel), exact-EV decisions on canonical hands (made pays ride, three-to-a-royal rides, junk pulls, four-flush rides at stage 2, low pairs pull), the frozen exact figures vs the published 3.51%/2.85%, payout mechanics (riding units only), determinism, and a 150,000-hand continuous-stream convergence. The slow test re-runs the full ~52M-state backward induction and must reproduce the frozen figures bit-for-bit.
- Last reviewed
- Verified by exact enumeration end-to-end (the strongest QC of the carnival games; no Monte-Carlo-only or cited-only figures anywhere). Matches the published Wizard-of-Odds values: 3.51% per base bet, ~2.85% element of risk. Last reviewed 2026-07-13.