He Took the Dealer’s Bust Card
It is the most-said sentence in the game. Somebody at third base hits when the chart says stand, the dealer turns over a winner, and the whole table knows exactly whose fault it was. This page answers it with the only two things that can settle it: a proof, and a measurement — and they answer different halves of the question.
The half that is a proof
Your expectation cannot change, and not because the effects are small or because they average out over a long enough session. They are identically zero, for a reason that needs no arithmetic at all.
A shuffled shoe has no memory and no order you know anything about. The card sitting second is as likely to be any particular card as the one sitting first. So when the player at third base takes the top card, the card the dealer now draws comes from exactly the same distribution as the one she would have drawn if he had not. Nothing about your hand’s prospects has moved. That is the whole argument, and it is a statement about shuffled decks rather than about blackjack.
This page deliberately does not publish a measured figure for that effect. A first attempt did, and got 0.000001 of a bet — which was not a small effect, it was the rounding error of a model that holds the shoe composition fixed mid-hand and therefore cannot represent a card leaving it at all. Reporting that number would have dressed an artifact up as a finding. The dataset behind this page uses a dealer recursion that removes each card for real, and burning one card before she draws moves her outcome distribution by 1.7 × 10⁻¹⁶ — zero, to the last bit a computer has. That is the proof reproduced, not a measurement of anything.
The half that is worth measuring
None of which tells you what that card would have done, and that is the interesting question, because the folklore has it half right. The card the third baseman took is precisely the card the dealer would have drawn first, if she drew at all — so its fate is exactly computable, and it has four possible endings rather than the two the complaint knows about.
- Busts her
- Makes her hand
- She draws again
- She never draws
What that one card would have done to the dealer · 6 decks · Dealer Stands on Soft 17, DAS, pays 3:2 · given no dealer blackjack
Take the up-card the argument almost always happens over. Against a 6, that card would have busted the dealer 31.9% of the time — which is exactly why the complaint feels so well-founded, because a third of the time the man really did take the bust card. But it would have made her hand 40.9% of the time, which is more often. Nobody ever says thank you for that, because it is invisible: when he saves the table, the hand simply happens and no one has anything to explain.
The other rows are stranger. Against an ace the card he took could never have busted her — 0%, not a rounding — because a dealer showing an ace who has already checked for blackjack is holding a soft hand, and a soft hand cannot bust on one card. Against a ten she does not even draw 58.2% of the time, so most of the time the card he took was never going to reach her at all. And against a 2, 3, 4 or 5 she is forced to take a card every single time, because no two-card hand starting that low can already be standing.
Across every up-card together, that card busts the dealer 18.2% of the time and never reaches her at all 31.7% of the time.
And she might have busted anyway
There is one more thing buried in “he took her bust card”: the assumption that she would not simply have busted two cards later instead. Against a 6 the dealer busts 42.3% of the time in total, and 31.9% of that is on the first card — so even at the up-card where the first card matters most, a quarter of her busts were going to come from somewhere else. The card he took was one draw in a sequence, not the hand.
What another player really costs you
There is a real cost to a crowded table, and it is nothing to do with anybody’s cards. The house edge is charged per hand, but you experience it per hour, so the number of rounds dealt is the number that actually decides what an evening costs.
| Players at the table | Rounds per hour | Units lost per hour | At $25 a hand |
|---|---|---|---|
| 1 | 209 | 0.95 | $23.87 |
| 2 | 139 | 0.64 | $15.88 |
| 3 | 105 | 0.48 | $11.99 |
| 4 | 84 | 0.38 | $9.59 |
| 5 | 70 | 0.32 | $7.99 |
| 6 | 60 | 0.27 | $6.85 |
| 7 | 52 | 0.24 | $5.94 |
At a 0.4568% house edge, flat bets · rounds per hour from Casino Operations Management (Kilby, Fox & Lucas)
Which inverts the complaint completely. A slow player at a full table is not costing you money — he is saving you money, at the rate of 4.0×, because every round he slows down is a round of house edge you never pay. The only way that flips is if your expectation is positive, and for anyone playing basic strategy it is not. The empty table you were hoping for is the same game charged four times as fast.
So what is left of it
The honest version is narrower than the complaint and more interesting than a flat denial. His play cannot move your expectation — that one is settled before you deal a card. It genuinely did cost the table that hand about a third of the time against a weak up-card, and genuinely did save it more often than that, and you only ever notice one of those. And the thing he actually changes, rounds per hour, works the opposite way from how the table assumes.
If you want the decisions that are genuinely close — where the person hitting really might have it right — those are the 31 contested spots, and what each one costs turns out to be startlingly little.