The blackjack house edge is the smallest in any standard casino game — between 0.4% and 0.6% for a player who follows basic strategy correctly, depending on rule variant. That number is everywhere on the internet, but it's rarely explained: where does it come from, what assumptions does it hide, and how much does it actually cost you per hour? This guide unpacks the math and gives you the realistic numbers.
What "house edge" actually measures
House edge is the casino's long-run profit as a percentage of total amount wagered. For blackjack, the accepted definition is:
house edge = -E[profit per hand] / average bet
E.g. if across millions of $1-base hands you lose an average of $0.005 per hand (counting all the doubles and splits that bump the actual wager above $1), the house edge is 0.5%. The denominator is "initial bet", not "total wager" — a doubled or split hand still counts as one original-bet unit in the standard convention.
A subtlety: house edge is the average, not the median. Blackjack distributions have heavy tails — most sessions you end either up or down a lot, not down 0.5%. The 0.5% number is what you converge to over thousands of hands.
How the 0.5% number is derived
For every starting hand × dealer upcard combination, you compute the expected value of the optimal action over the full distribution of dealer outcomes (given the cards already used up). Then weight each combination by its probability of occurring at the start of a hand. Sum.
Concretely, for a 6-deck S17 DAS no-surrender game, the weighted sum lands at roughly:
- Blackjack (you): +0.7% (1/21 of hands × 1.5× payout × win rate)
- Doubling: +0.2%
- Splitting: +0.1%
- Straight-play losses: -1.5%
- Net: ≈ -0.5%
Those component numbers are approximate. The precise breakdown depends on how you draw the lines around what counts as "doubling" versus "splitting", but the net figure is robust to fractions-of-a-tenth-of-a-percent precision under any reasonable accounting.
How rule variations move the number
The 0.5% baseline is for one specific ruleset. Real casinos deviate. Here's how the most common rule changes shift the house edge — each line shows the change vs the 6-deck S17 DAS no-surrender baseline:
| Variation | Edge change |
|---|---|
| Single deck (vs 6 decks) | −0.50% (better for player) |
| Dealer hits soft 17 (H17) | +0.22% (worse) |
| Blackjack pays 6:5 (vs 3:2) | +1.40% (catastrophic) |
| Surrender allowed (late) | −0.07% (better) |
| Double after split (DAS) | −0.13% (better) |
| No DAS | +0.13% (worse than baseline) |
| European no-hole-card rule | +0.11% (worse) |
Per-hour expected loss
House edge per hand × hands per hour × average bet = expected loss per hour. Realistic ranges:
- Live table (Evolution): ~60 hands/hr, $10 average bet, 0.5% edge → ~$3/hr expected loss.
- Live solo / fast table: ~100 hands/hr → ~$5/hr.
- Online RNG (no dealer wait): can hit 300 hands/hr → ~$15/hr. Volume matters.
Important: these are expected values over many hours. Any specific hour-long session has a standard deviation of roughly ±$110at $10/hand. Variance dominates expected loss until you've played a lot of hands.
Variance: what to actually expect
Blackjack's per-hand variance is around 1.3 (in units-squared). The standard deviation of a session of n hands at $1 average bet is roughly √(1.3 × n) — at 60 hands, that's ~$8.80 of typical swing. At $10/hand it scales to ~$88.
Concretely: in a one-hour session of 60 hands at $10/hand, with 0.5% house edge:
- Expected outcome: −$3
- 1 SD range: −$91 to +$85 (covers ~68% of sessions)
- 2 SD range: −$179 to +$173 (covers ~95% of sessions)
You will have winning hours, even with a 0.5% disadvantage — about 47% of one-hour sessions end up positive under these parameters. Over 1,000 hours of play, the variance averages out and the long-run loss converges to the expected −$3/hr × 1,000 hr = $3,000.
Where each game stands
Blackjack's ~0.5% edge is unique among standard casino games. For context:
| Game | Typical house edge |
|---|---|
| Blackjack (basic strategy) | 0.4–0.6% |
| Baccarat (banker bet) | 1.06% |
| Craps (pass/come) | 1.41% |
| European roulette | 2.70% |
| American roulette | 5.26% |
| Crash games (Stake, Roobet) | 1.0–3.0% |
| Slots (mainstream) | 3–10% |
| Keno / wheel side bets | 10–25% |
Blackjack is mathematically the cheapest game in the room. That's only true if you play it well — at ~5% house edge for untrained play, it's squarely in slot territory.
Practical bankroll math
How much should you bring to a session? A useful rule of thumb: bring enough that ruin-by-variance is unlikely over the session length. For a 0.5% game at $10/hand:
- 1-hour session: $200 bankroll has about a 1% chance of ruin (you go all-in down).
- 4-hour session: $400 bankroll → ~3% ruin chance.
- Doubling your bet on losses (Martingale): inflates short-term win rate but explodes ruin risk — don't.
The doubling-on-loss systems are mathematical illusions: their expected return is exactly the same as flat betting (no strategy changes house edge), but their variance is much worse. They feel like winning because most sessions end positive — until the one big loss wipes everything.
What about side bets?
Side bets — 21+3, Perfect Pairs, Lucky Ladies, Bust It, Insurance — all carry house edges in the 4-15% range. They're marketed alongside the main game but priced like slot side bets. The strategy & side bets pagehas the full ordered table — house edge ascending, so the "least bad" sit at the top. Even the best side bet is 8× the house edge of the main game; the worst are 30×.
The right way to think about side bets: they're lottery tickets sold at the blackjack table. If you want occasional big-multiplier excitement, fine — but they're strictly an EV drain on top of your main play.
Putting it together
If you want the cheapest entertainment a casino offers, play blackjack with basic strategy at a 3:2 table with S17 and DAS — pretty much the default at live tables — and skip the side bets and the insurance. That's a ~0.5% house edge, an expected $3/hr loss at $10/hand, and the lowest long-run cost-per-hour of any traditional casino game.
If you want to lower the edge even further: the only legal path is card counting, which works in principle on the right shoe-style tables but is impractical on most online live-dealer tables (where reshuffles and bet-size flags defeat it).
To practise execution under real conditions, the free simulator tracks your strategy accuracy and shows the long-run EV cost of any deviations after every hand. The basic strategy guide covers the chart cell-by-cell with the reasoning behind each decision; the how-to-play explaineris the starting point if you're new to the game.
