Games · Blackjack
Blackjack: see the real odds
Blackjack has the lowest house edge in the casino — if you play it right. Set the rules below, run thousands of hands on computer-perfect basic strategy, and watch the simulated edge track the analytical theory line — exact in unlimited mode, a fresh-shoe reference for finite decks. Then look at the spread: even flawless play swings up and down.
entertainment only
Blackjack simulator
Pick the rules, run thousands of hands on perfect basic strategy, and see how the edge, variance, and long run play out.
⚠ Settings changed — run the simulation again to update the results below.
- Initial rounds
- 0
- Sessions
- 0
- Avg session
- —
- Std dev
- —
- Best / worst
- —
- Resolved W/P/L
- —
Theory is exact for the unlimited deck, a close estimate for finite decks — methodology & assumptions
Fixed assumptions: US hole-card peek · split once (max 2 hands) · no resplitting aces · split aces draw one card · insurance always declined · flat 1-unit initial bets · finite shoes reshuffle at the cut card — 75% penetration for 2+ decks, and earlier on a single deck (≈50%) so a full round always fits.
Edge is per initial bet. Doubles and splits put more chips in play, so loss per $100k of total action is a little lower. Run a simulation to measure the action multiplier.
Finite decks: the theory line is a fresh-shoe reference, not the convergence target. The theory line is basic strategy's edge on a fresh shoe — exact for the unlimited / independent-draw model, but not a rigorously exact finite calculator (finite hit/double/split EV after a dealer peek is approximated, and finite splits are treated as independent draws). The simulation deals a depleting shoe, which is a genuinely different process, so the observed edge settles above the fresh-shoe line by the cut-card effect: tiny for 6–8 decks (≈0.01–0.02pp), but sizable for 1–2 decks (≈0.1pp) — large enough that a one-deck run can land on the other side of break-even from the fresh-shoe line. So for small decks, treat the line as a reference value, not the number the simulation "converges to."
Distribution. If the hand count doesn't divide evenly by session length, the final partial session is excluded from the spread.
Perfect basic strategy · entertainment & education only — no real money.
How this simulation works
- Rules modeled
- Configurable: 1–8 decks or unlimited; dealer stands/hits soft 17; double after split; blackjack pays 3:2 or 6:5; late surrender; double-down restrictions. Fixed this build: US hole-card peek, split once (max 2 hands), split aces draw one card, insurance always declined; finite shoes reshuffle at the cut card (75% for 2+ decks, ≈50% on a single deck so a full round always fits).
- Assumptions
- Flat 1-unit bets. The simulated player follows computer-perfect basic strategy derived for the rules selected. Unlimited mode draws cards with replacement, so rounds are independent; finite mode deals consecutive rounds from a shuffled, depleting shoe until the cut card, so rounds within a shoe are not independent.
- Mathematical basis
- The theory line is computed analytically (dealer outcome distribution → player expected value): exact for the unlimited (independent-draw) model the default uses, and a close analytical fresh-shoe estimate for finite decks. The Monte Carlo simulation is an independent check — the unlimited model converges to the line, while finite decks settle a bit away from it (the cut-card effect: small for 6–8 decks, larger for 1–2).
- Engine version
- blackjack 0.4 · rules artifact 2026-06-21
- Validation
- 137 fast regression tests, 4 high-sample validation tests, and a regenerated 240-rule finite-edge table; last independently reviewed 2026-06-21 (5 review passes).
- Last reviewed
- 2026-06-21