Blackjack has the reputation of being the “beatable” casino game. For almost every player it is not, but it is the game where your own decisions move the cost the most. To see why, it helps to know exactly where the casino’s advantage comes from, which rules change it, and which of your habits quietly add to it.
This guide is about blackjack only. If you want to see how blackjack compares with other games, the wider comparison covers that. Here the question is narrower: why is there an edge at all, and what moves it?
The one rule that creates the blackjack house edge
The dealer and the player play almost the same game. Both try to get closer to 21 without going over. The difference is the order: you act first. If you bust, your bet is taken immediately. The dealer never has to play a hand against yours, so a dealer bust later in the round does not save you.
That single asymmetry is the source of the house edge. Under the engine’s default rules, with six decks and the dealer standing on soft 17, the dealer busts in 28.2% of rounds. That sounds like a large advantage for the player. But every time you have already busted, the dealer’s bust does not matter, and the house keeps the bet.
Notice what the edge is not. It is not a hidden rule that changes the cards, and it is not a property of “hot” or “cold” tables. The cards are a fair random draw. The edge lives in the order of play, and in the payouts attached to each outcome.
Why the dealer’s bust chance depends on the upcard
The dealer has no choices. The dealer draws to a fixed rule, so the distribution of final hands depends only on the upcard. This is what lets a player’s strategy exist at all. Our exact dealer table, computed by the engine for six decks and a stand on soft 17, shows how sharply the bust chance varies:
| Dealer upcard | Dealer busts |
|---|---|
| 2 | 35.4% |
| 3 | 37.4% |
| 4 | 39.6% |
| 5 | 41.8% |
| 6 | 42.3% |
| 7 | 26.2% |
| 8 | 24.4% |
| 9 | 22.9% |
| 10 | 21.2% |
| A | 11.5% |
The ten and ace rows include the cases where the dealer has a natural and therefore does not draw.
Source: Our blackjack odds study (S5), exact engine, 2026-10-09
A dealer showing a 6 busts about 42% of the time. A dealer showing a 10 busts about 21% of the time, and a dealer showing an ace busts about 12% of the time. Basic strategy is largely a way of using these numbers: stand on stiff hands when the dealer is likely to bust, and go after the hands that can win more when the dealer is likely to finish with a decent total.
Where you win it back
Every option the dealer does not have is a way to recover part of the edge the first-to-act rule gives away:
- The 3:2 blackjack payout. You are paid 1.5 times your bet for a natural. A dealer natural only takes your one unit.
- Doubling down. You can double your bet when you are the favourite, such as 11 against a dealer 6.
- Splitting pairs. Splitting 8s turns one bad hand into two playable ones. Splitting aces gives you two strong starts.
- Standing on stiff hands. You can stand on 12 to 16 when the dealer shows a weak card and is likely to bust. The dealer is forced to keep hitting.
- Surrender, where offered, lets you give up half the bet on your worst hands.
Use all of these correctly, following a basic strategy chart for your exact rules, and the cost falls to one of the lowest in the building. Our engine puts the house edge of the default six-deck game (stand on soft 17, double after split, late surrender, split up to four hands, hole-card peek, 3:2) at 0.33%. Remove late surrender and keep everything else the same, and the exact figure is 0.4%. The two numbers are close, but they are different games, and a rule you may not even notice on the felt accounts for the gap.
It is still not zero. The edge is small, but it does not go away.
The blackjack house edge is a set of small rule effects
Because the total is so small, the printed rules matter. Each rule on the felt nudges the edge up or down by a few hundredths of a percentage point, with one big exception. The table below compares each change against the engine’s default rules. “Published” is the effect reported by an independent published analysis, and “computed” is the change in the edge from this site’s own exact engine. They agree closely in most rows and disagree a little in a few, which is why both are shown.
| Rule change from the default game | Published | Computed |
|---|---|---|
| Blackjack pays 6:5 instead of 3:2 | 1.39 | 1.36 |
| Dealer hits soft 17 (H17) | 0.22 | 0.20 |
| Doubling only on 10 and 11 | 0.18 | 0.19 |
| Doubling only on 9, 10 and 11 | 0.09 | 0.10 |
| No doubling after splitting | 0.14 | 0.14 |
| No late surrender | 0.07 | 0.07 |
| No hole card (European) | 0.11 | 0.11 |
| Only one split allowed (two hands) | 0.10 | 0.05 |
| Aces may be resplit | −0.08 | −0.07 |
| Eight decks instead of six | 0.02 | 0.02 |
| Two decks instead of six | −0.17 | −0.19 |
| One deck instead of six | −0.46 | −0.51 |
Rows with a negative sign help the player. Computed values are from the site's blackjack engine; published values come from the Wizard of Odds rule-variation tables, as of 2026-10-08.
Source: Wizard of Odds rule variations; site blackjack engine
Source: Wizard of Odds rule variations (2026-10-08), via the site's blackjack engine
See the numbers as a table
| Value | |
|---|---|
| Blackjack pays 6:5 instead of 3:2 | 1.39 pts |
| Dealer hits soft 17 (H17) | 0.22 pts |
| Doubling only on 10 and 11 | 0.18 pts |
| Doubling only on 9, 10 and 11 | 0.09 pts |
| No doubling after splitting | 0.14 pts |
| No late surrender | 0.07 pts |
| No hole card (European) | 0.11 pts |
| Only one split allowed (two hands) | 0.10 pts |
| Aces may be resplit | −0.08 pts |
| Eight decks instead of six | 0.02 pts |
| Two decks instead of six | −0.17 pts |
| One deck instead of six | −0.46 pts |
Two things stand out.
First, 6:5 dwarfs everything else. Paying $6 on a $5 natural instead of $7.50 adds about 1.39 percentage points by the published figure, and about 1.36 by the engine. The next largest change on the list is a fraction of that. A single-deck game is valuable only when it still pays 3:2, which is why a one-deck table that pays 6:5 is usually a worse game than a six-deck table that pays 3:2.
Second, the small rules add up. Dealer hits soft 17, no double after split and no surrender each look minor. Stack them and the cost of a table can more than double. The engine’s baseline of 0.33% becomes 0.76% once these three are combined, which is why it pays to read the rules before you sit down.
When published and computed figures differ
Where the two columns differ, the computed number is the one from this site’s own calculation, and the published number is the one in the independent reference. They can differ for a few reasons. The published figures are rounded to hundredths. Rules interact: the cost of losing surrender depends on whether the dealer peeks, and the benefit of resplitting aces depends on the number of decks. The engine also uses a documented approximation for the player’s later draws, and a “split once” restriction is one place where the engine and the reference are further apart than usual (about 0.05 against 0.10 percentage points). If a figure matters to a decision, treat the pair as a range rather than a point.
What the blackjack house edge means in money
An edge is a percentage of the money you put through the game, not of your bankroll. It helps to turn it into dollars per hour, and the cost per hour depends heavily on pace.
At a $10 average bet under the six-deck stand-on-soft-17 rules without surrender (0.4% edge), the study’s central pace of 70 hands an hour gives an expected loss of $2.80 an hour. The slow-to-fast band of 50 to 120 hands an hour gives $2.00 to $4.80. At $25 a hand the central figure is $7.00 an hour, and at $100 a hand it is $28.00.
Those hands-per-hour figures are assumptions, not measurements. A full table with six players deals fewer hands than a table with one player, and an online live table deals faster than either. The honest way to read the numbers is as a band.
Expected loss is not a forecast
The figures above are long-run averages under perfect basic strategy. Any one hour can finish well ahead or well behind. The expected loss only tells you what an hour costs on average, over many hours.
Numbers that look like the edge but are not
Several well-known numbers get confused with the house edge, so it helps to separate them.
The chance of being dealt a blackjack is about 4.75% at six decks, or 1 in 21.1. With a single deck it is 4.83%. The deck count changes the natural’s frequency only a little. What changes the cost is what the natural pays: 3:2 or 6:5.
How often a basic-strategy player wins is less than half of rounds. In the study’s simulation of twenty million six-deck rounds, the player wins 42.4% of rounds, loses 49.1% and pushes 8.5%. How can a game with more losses than wins cost only a fraction of a percent? Because wins are worth more than losses: blackjacks pay 3:2, and doubles and splits put more money on the table when the odds favour the player.
Losing streaks are normal. With those frequencies, treating each round as independent, a first losing streak of seven hands arrives after about 284 hands on average. Over 1,000 hands there is about an 82% chance of at least one streak of eight losing hands. None of this means the game is rigged or that you are doing something wrong. It is the ordinary shape of a game with a small edge. If you want a deeper look at the probabilities, the planned odds guide will cover them in full.
Insurance: the bet that raises the cost
When the dealer shows an ace you are offered insurance, a side bet of up to half your stake that the dealer has a ten underneath. It pays 2:1. The name suggests protection, but it is a separate bet with its own edge.
Under the engine’s six-deck rules the insurance bet has a house edge of 7.4% of the amount insured, which matches the published figure. Because the insurance stake is half your main bet, the cost per round when it is taken is about 3.7% of the main bet. The situation arises about once every 13 hands. Taking insurance every time it is offered would cost about 0.28 units per 100 hands, which is the highest figure in the next section. “Even money” on a natural against an ace is the same bet in disguise.
Counting the tens in your own hand shifts the figure slightly. Holding tens makes insurance a little worse, and holding none makes it a little better, but at an ordinary table the right answer is to decline.
What common mistakes cost
The edge figures above assume every decision is right. Real players make errors, and the engine can price each one. The table below lists twelve common mistakes under the default six-deck rules. “Cost per hand” is the expected loss in units of the original bet each time the mistake is made, measured against the best legal play. “Once in” is how often the player reaches that exact decision with no dealer natural, and the last column is what the habit costs over 100 hands if you made it every time.
| Mistake | Cost per hand (units) | Once in (hands) | Cost per 100 hands (units) |
|---|---|---|---|
| Not splitting A-A against a dealer 6 (hitting soft 12) | 0.492 | 2,271 | 0.022 |
| Splitting 10-10 against a dealer 6 | 0.251 | 137 | 0.183 |
| Standing on soft 18 (A-7) against a dealer 9 | 0.083 | 1,088 | 0.008 |
| Standing on hard 16 against a dealer 7 | 0.068 | 219 | 0.031 |
| Not splitting 8-8 against a dealer 10 (standing on 16) | 0.061 | 615 | 0.010 |
| Hitting instead of doubling 11 against a dealer 10 | 0.060 | 74 | 0.081 |
| Standing on hard 16 against a dealer 10 | 0.040 | 59 | 0.067 |
| Taking insurance (or even money) against a dealer ace | 0.037 | 13 | 0.284 |
| Hitting instead of doubling 10 against a dealer 9 | 0.032 | 363 | 0.009 |
| Standing instead of doubling soft 18 (A-7) against a dealer 3 | 0.028 | 1,088 | 0.003 |
| Standing on hard 12 against a dealer 3 | 0.018 | 156 | 0.012 |
| Hitting hard 13 against a dealer 2 | 0.015 | 155 | 0.010 |
Insurance costs are per hand of the main bet: the insurance stake is half the main bet. A unit is the original bet.
Source: Our blackjack odds study (S5), exact engine, 2026-10-09
Source: Our blackjack odds study (S5), exact engine, 2026-10-09
See the numbers as a table
| Value | |
|---|---|
| Not splitting A-A against a dealer 6 (hitting soft 12) | 0.02 |
| Splitting 10-10 against a dealer 6 | 0.18 |
| Standing on soft 18 (A-7) against a dealer 9 | 0.01 |
| Standing on hard 16 against a dealer 7 | 0.03 |
| Not splitting 8-8 against a dealer 10 (standing on 16) | 0.01 |
| Hitting instead of doubling 11 against a dealer 10 | 0.08 |
| Standing on hard 16 against a dealer 10 | 0.07 |
| Taking insurance (or even money) against a dealer ace | 0.28 |
| Hitting instead of doubling 10 against a dealer 9 | 0.01 |
| Standing instead of doubling soft 18 (A-7) against a dealer 3 | 0.00 |
| Standing on hard 12 against a dealer 3 | 0.01 |
| Hitting hard 13 against a dealer 2 | 0.01 |
The ranking is not the order of the cost per hand. Not splitting aces against a dealer 6 is the single most expensive error in the table per occurrence, at about 0.49 units, but it comes up only about once in 2,300 hands. Splitting tens is also costly per occurrence, and tempting, but it is also rare. Standing on hard 16 against a dealer 10 costs less per occurrence, about 0.04 units, but it is common, about once every 59 hands. Hitting instead of doubling an 11 against a 10 is similar. The frequent small leaks matter most over a session, and the expensive rare ones matter most when they do come up.
The number that matters for your wallet is the last column. Over a few hundred hands the biggest leak in the table is insurance, because it is offered so often. Next comes splitting tens, because two hands feel better than one, then hitting an 11 instead of doubling it, then standing on 16 against a ten.
How to read a table before you sit down
You do not need to memorise the table above. A short checklist covers most of the cost of the rules:
- Blackjack payout. Look for 3:2. If it says 6:5, find another table.
- Dealer on soft 17. Stand is better for the player. The difference is about a fifth of a percentage point.
- Doubling after a split and surrender. Both help the player, and both are worth looking for.
- Doubling restrictions. “Double on 10 and 11 only” costs more than most players expect.
- Number of decks. Fewer is better, but only if the payout and the other rules are intact.
Online versions can differ from the felt. Read the rules page of the specific game you are about to play, because live-dealer and RNG versions of the same game are not always set up the same way.
The honest summary
Basic strategy does not beat blackjack. It lowers the price of playing to one of the lowest on a casino floor. No betting system, progression or “hot table” changes that arithmetic, because none of them change who has to act first, and none of them change the payout on a natural.
What you can control is short. Choose a 3:2 table, read the rest of the rules, play the chart for those rules, decline insurance, and decide the budget before the first hand. A cheap game is still a game with a cost. If it stops being something you can afford, free, confidential help is available at the link at the foot of this page.
Related guides
- Full basic strategy chart: every play for a stand-on-soft-17 game, cell by cell.
- Basic strategy when the dealer hits soft 17: the handful of plays that change under H17, and what the rule costs.
- One deck at the table: what changes: how one deck shifts the chart, and why 6:5 payouts cancel the benefit.
- How blackjack compares with the rest of the casino: about 30 bets in order, with cost per hour.