PRO
More systems, strategy, and simulation
Save 25% annually
$9.99 USD/month
$7.49 USD/month*
Annual billing: $89.91 per year.
Original data
What three common blackjack mistakes actually cost
Four million hands per scenario through the BJCPRO engine, to put a number on the three errors our basic strategy guide warns about.
Most players assume insurance is the expensive habit. It is not. Splitting a pair of tens gives away nineteen times more expected value, and playing soft hands as if they were hard costs nearly three times what insurance does.
The short answer
Splitting tens is the costly mistake, and it is not close
Measured against a correct basic-strategy baseline, always taking insurance costs about 0.30 percentage points per hand. Splitting a pair of tens costs about 5.79 — roughly nineteen times more. Playing soft hands as if they were hard sits between them at 0.87.
These are averages per hand played, in units of your base bet. They describe how much expected value each habit gives away over the long run, not what will happen in any single session.
The numbers
Cost of each mistake, four million hands each
Six decks, dealer stands on soft 17, blackjack pays 3:2, 75% penetration, double after split allowed, late surrender allowed.
| Scenario | EV per hand | Cost vs correct play |
|---|---|---|
| Basic strategy, played correctly | −0.512% | baseline |
| Soft hands played as if hard | −1.381% | 0.87 points |
| Insurance taken every time it is offered | −0.814% | 0.30 points |
| Pairs of tens split | −6.302% | 5.79 points |
Margin of error ±0.113 points at 95% confidence.
Why tens are so costly
You are breaking up the best hand in the game
A pair of tens is a 20. It wins outright most of the time and only loses to a dealer 21. Splitting it trades one very strong hand for two ordinary ones and doubles the money at risk while doing so. The size of the loss surprises people because it compounds: the simulation allows resplitting up to three times, exactly as the product does, so one bad decision can turn a single strong hand into four mediocre ones in the same round.
- A pair of tens shows up in roughly one hand in ten, so the mistake is not rare.
- Standing on 20 is one of the most profitable positions in the game, and splitting gives it away voluntarily.
- The loss compounds with every resplit, which is why it dwarfs the other two errors.
The insurance surprise
Insurance is a bad bet, but a small one
Insurance gets more attention than its cost deserves, and the arithmetic is simple enough to check by hand. With six decks the dealer holds a ten under an ace about 31% of the time, so a half-unit insurance bet returns about −0.034 units. An ace shows roughly one hand in thirteen. Multiply those and you get about 0.26 points per hand, close to the 0.30 the simulation measured — a useful sign that the model behaves.
It is still a bet to decline without a count that justifies it. It is simply not the habit costing you the most money.
Soft hands
Treating A-7 like a hard 18 throws away free chances
A soft hand cannot bust on the next card, which is exactly what makes it worth playing differently. A player who memorises totals without separating soft from hard stands on soft 18 against a dealer 9 or 10, when the correct play is to keep drawing. That single confusion costs about 0.87 points per hand — nearly three times the insurance habit — and it comes from a gap in understanding rather than a deliberate gamble.
How this was measured
Reproducible, and worth checking
The simulation reuses the product rather than reimplementing it. Hand evaluation comes from the same pure domain code that runs the BJCPRO tables, and the strategy is the exact matrix the trainer uses, with the same row mapping. The shuffle uses a fixed seed, so anyone running the script gets the same figures.
- Four million hands per scenario, one scenario per mistake, against the same seeded shoe sequence.
- Each variant changes exactly one decision, so the difference isolates that habit.
- Split aces receive one card and stand, and aces cannot be resplit, matching the engine.
An honest limitation
Our baseline lands slightly worse than the 0.40–0.46% usually published for these rules, because a few conditional cells resolve more simply here than in the product. That makes the differences more trustworthy than the absolute numbers: a small systematic bias in the baseline largely cancels when you subtract it. Treat "splitting tens costs 5.79 points" as the finding, not "the house edge is exactly 0.512%".
Practise the hands
Numbers change behaviour only if you drill them
Knowing that splitting tens is expensive does not stop the impulse at the table. The three hands in this article are worth practising until the correct play is automatic.
Frequently asked questions
Does this mean I will lose 0.512% of my money?
No. It is an average across four million hands. Individual sessions swing far wider in both directions, which is what variance means.
Would card counting change these numbers?
For insurance, yes: it becomes correct at a high enough true count. Splitting tens and misplaying soft hands stay wrong under any normal count.
Can I reproduce these figures?
Yes. The simulation script uses a fixed seed, so the same command returns the same numbers.
Is BJCPRO telling me how to bet?
No. BJCPRO is a training and analysis tool. It does not take bets, operate a casino, or promise outcomes.
Sources and further reading
- Wizard of Odds — Blackjack house edge calculator — Reference figures per rule set, used to sanity-check the baseline.
- Edward O. Thorp, Beat the Dealer — The origin of basic strategy as a computed result rather than folklore.
- Peter Griffin, The Theory of Blackjack — Expected value and the cost of departing from correct play.
- National Council on Problem Gambling — Responsible play resources.