- Rules modeled
- Standard Four Card Poker, single 52-card deck, independent hands. The player antes and receives 5 cards; the dealer receives 6 (one exposed). The player folds (forfeiting the ante) or raises a Play bet of 1× to 3× the ante. Both sides make their best FOUR-card poker hand (ranking: four of a kind > straight flush > three of a kind > flush > straight > two pair > pair > high card; ace high or low in straights). There is no dealer qualifier; the higher hand wins even money on Ante and Play, and ties go to the player. The Ante Bonus pays on the ante regardless of the dealer: three of a kind 2:1, straight flush 20:1, four of a kind 25:1. The optional Aces Up side bet pays on the player's own best 4-card hand (pair of aces or better; 50-40-8-5-4-3-1 table). Stakes are fun credits in integer cents.
- Assumptions
- Hands are independent (a freshly shuffled deck each round). The simulator plays the published SIMPLE strategy: raise 3× with a pair of tens or better, 1× with a pair of twos through nines, fold otherwise, which ignores the dealer's exposed card. Expert play that uses the up card reaches a slightly lower edge (shown in the Odds tab). Fun credits only, no real money.
- Mathematical basis
- The Aces Up side bet and the hand-frequency table are EXACT, enumerated over all C(52,5)=2,598,960 best-4-of-5 hands (edge 3.89% on the 50-40-8-5-4-3-1 table), and the fold/1×/3× play distribution is exact from the same enumeration. The Ante-Play house edge cannot be exactly enumerated (the joint player×dealer space is ~10¹³ deals), so the engine carries the published Wizard-of-Odds figures, 3.40% of the ante for the simple strategy (2.79% optimal), verified by seeded Monte-Carlo convergence in the Lab. The exact and published figures are labeled as such throughout.
- Engine version
- four-card-poker engine 1.0
- Validation
- 15 unit tests + an opt-in slow test; the 4-card evaluator across all categories including the trips-beat-flush and flush-beats-straight quirks and the A-2-3-4 wheel; best-4-of-5 and best-4-of-6 selection; the simple-strategy thresholds; the exact Aces Up counts (624 quads / 2,072 straight flushes / 81,096 ace pairs of 2,598,960) and its 3.89% published edge to 4dp; the Ante Bonus schedule; seeded determinism; and a 500,000-hand convergence toward the published 3.40%. The slow test re-runs the full 2.6M enumeration and must reproduce the frozen counts exactly.
- Last reviewed
- Verified by exact enumeration where the math allows (Aces Up + frequencies + play distribution) and against the published Wizard-of-Odds figures where it doesn't (Ante-Play, MC-converged). Building this engine also uncovered and fixed a systematic Monte-Carlo bias (per-hand RNG re-seeding), validated against Three Card Poker's exactly-known edge. Last reviewed 2026-07-13.
21+