What Bad Blackjack Costs
The close-call pages measure how narrowly the correct play wins a decision. This page measures something different and more useful: what getting that decision wrong every single time actually costs you, across every hand you will ever be dealt.
Those are not the same quantity. A margin is per-decision. A cost is per-decision multiplied by how often the decision comes up — and only the second one is denominated in money. Measuring it inverts the whole subject.
The habits
Each of these is a rule somebody at the table genuinely plays by. Each was turned into a policy and the entire game re-solved under it, against a baseline of perfect basic strategy at 0.4568% — 6 decks, dealer standing on soft 17, double after split allowed, naturals paying 3:2.
Points added to the house edge · 6 decks · Dealer Stands on Soft 17, DAS, pays 3:2
Note where the two hard-16 habits land. Refusing ever to hit a 16 costs 0.2539 points; always hitting it costs 0.736 — nearly three times as much. The hand has a reputation as the one you can never get right, and the truth is narrower: the stubborn version of the mistake is much cheaper than the reckless one.
Why playing like the dealer loses
The most expensive habit on the list is also the most reasonable-sounding: play exactly the way the dealer plays, and surely you break even. You do not, and the reason decomposes exactly.
Strip out every advantage the player has — pay a natural at even money instead of 3:2 — and a player who mimics the dealer faces a house edge of 7.9355%. That figure is not an accident of the engine: it is the probability that both hands bust, computed separately as 7.9363%. When you bust and the dealer then busts too, you have already paid. That single asymmetry — the player acts first — is the entire house edge in a game played symmetrically.
Handing back the 3:2 on a natural is worth 2.2661 points, which is why the real mimic-the-dealer figure lands at 5.6694%. Commonly published figures put it near 5.48%; ours is a little higher because the game modelled here is a little different, and the identity above is the check that matters, since both sides of it are computed here independently.
The arguments, priced
Now the 31 decisions the close-call pages exist for — the ones that get argued about at real tables. Each row is what it costs to play that hand the wrong way every time you hold it against that up-card. The column on the right is the same figure in money, at $10 a hand over 100 hands.
| Hand | Instead of | Margin | Costs | Per 100 hands |
|---|---|---|---|---|
| Hard 15 v 10 | Stand ← Hit | 0.0360 | 0.1306 | $1.31 |
| Hard 16 v Ace | Stand ← Hit | 0.1498 | 0.0994 | 99 cents |
| Hard 16 v 9 | Stand ← Hit | 0.0338 | 0.0325 | 32 cents |
| Hard 12 v 6 | Hit ← Stand | 0.0168 | 0.0145 | 15 cents |
| Hard 13 v 2 | Hit ← Stand | 0.0150 | 0.0145 | 15 cents |
| Hard 12 v 3 | Stand ← Hit | 0.0186 | 0.0143 | 14 cents |
| Hard 16 v 10 | Stand ← Hit | 0.0006 | 0.0129 | 13 cents |
| Hard 9 v 3 | Hit ← Double | 0.0196 | 0.0076 | 8 cents |
| Hard 11 v Ace | Double ← Hit | 0.0339 | 0.0067 | 7 cents |
| Hard 12 v 4 | Hit ← Stand | 0.0025 | 0.0037 | 3.7 cents |
| Pair of 2s v 3 | Hit ← Split | 0.0570 | 0.0027 | 2.7 cents |
| Soft 13 v 6 | Hit ← Double | 0.0181 | 0.0024 | 2.4 cents |
| Soft 14 v 5 | Hit ← Double | 0.0141 | 0.0022 | 2.2 cents |
| Pair of 6s v 2 | Hit ← Split | 0.0348 | 0.0021 | 2.1 cents |
| Hard 9 v 2 | Double ← Hit | 0.0133 | 0.0021 | 2.1 cents |
| Pair of 3s v 3 | Hit ← Split | 0.0434 | 0.0020 | 2.0 cents |
| Soft 16 v 4 | Hit ← Double | 0.0185 | 0.0019 | 1.9 cents |
| Pair of 4s v 6 | Hit ← Split | 0.0310 | 0.0016 | 1.6 cents |
| Soft 16 v 3 | Double ← Hit | 0.0163 | 0.0014 | 1.4 cents |
| Soft 19 v 6 | Double ← Stand | 0.0164 | 0.0013 | 1.3 cents |
| Pair of 2s v 2 | Hit ← Split | 0.0260 | 0.0012 | 1.2 cents |
| Pair of 4s v 5 | Hit ← Split | 0.0095 | 0.0009 | 0.9 cents |
| Pair of 2s v 8 | Split ← Hit | 0.0148 | 0.0008 | 0.8 cents |
| Pair of 7s v 8 | Split ← Hit | 0.0171 | 0.0006 | 0.6 cents |
| Soft 18 v Ace | Stand ← Hit | 0.0073 | 0.0005 | 0.5 cents |
| Pair of 3s v 8 | Split ← Hit | 0.0124 | 0.0005 | 0.5 cents |
| Soft 17 v 2 | Double ← Hit | 0.0066 | 0.0004 | 0.4 cents |
| Pair of 3s v 2 | Hit ← Split | 0.0026 | 0.0003 | 0.3 cents |
| Soft 18 v 2 | Double ← Stand | 0.0020 | 0.0003 | 0.3 cents |
| Soft 15 v 4 | Double ← Hit | 0.0009 | — | nothing |
| Soft 13 v 5 | Double ← Hit | 0.0074 | — | nothing |
Read the bottom of that table, then read the chart again. Hard 16 against a ten — the single most argued-about hand in blackjack — costs 13 cents per 100 hands at $10 if you get it wrong every time for the rest of your life. Always splitting tens, which nobody argues about because everyone who does it is certain, costs 249× as much. The spots that generate arguments are almost exactly the spots where being wrong is free, and that is not a coincidence: a decision only becomes contentious when the two options are worth nearly the same, which is the definition of a cheap mistake.
The expensive one in that table is Hard 15 against 10 at 0.1306 points, which is 460× the cheapest — and it is expensive for a reason that has nothing to do with the decision being hard. You simply hold that hand a lot.
Two that are not mistakes at all
Soft 15 against 4 and Soft 13 against 5 cost exactly nothing, and the reason is worth more than the rows themselves. A basic strategy chart is computed against an infinite shoe. Deal a real six-deck game and remove the player’s own two cards and the dealer’s up-card, and a decision this close can land the other way — here both do, and the “mistake” is fractionally the better play. Both models are defensible; they only ever disagree where the margin is already a coin-flip, which is exactly where these sit.
What to take from this
Not that basic strategy is optional. Every figure above is a cost, none of them is negative, and the habits at the top of the chart turn a nearly-even game into one of the worst bets on the floor. What the numbers do say is where the attention belongs: not on the hand you and the person next to you disagree about, but on whether you are doubling and splitting when you should — those two alone are worth 2.00 points, more than every contested decision on this page combined.
And none of it changes the sign. A perfect basic strategy player still loses 0.4568% of everything they stake, slowly. How slowly, and how much it bounces around on the way, is on the variance page.