August 3, 2026 · 9 min read · blackjack · insurance · index plays

Blackjack insurance: a 7.4% bad bet, with one exception

No — insurance is a losing bet in blackjack unless you are counting cards, with a house edge near 5.9% in a single deck and 7.4% in a six-deck shoe. Those are not fractions of a percent. They are side-bet numbers, an order of magnitude above the cost of the game you sat down to play, and they are attached to the most reassuring word on the felt.

The exception is real, and not a technicality. Insurance is the one bet a card counter should sometimes make with enthusiasm, because it is the single decision the count predicts best. Two things have to be true at once: you have to be counting, and the count has to be high enough. Here is the arithmetic for both halves.

What you are actually betting on

When the dealer's upcard is an ace, you are offered insurance: a side bet, up to half your original wager, that pays 2 to 1 if the hole card is any ten-value card. Wikipedia's description is blunt and correct — it is a side bet that the dealer has a blackjack.

Read that again and notice what is missing. Your cards are not in it. Insurance does not protect your 20, it does not hedge your 12, and it does not care whether you are about to double. It is a proposition on one unseen card, settled before you make a single playing decision, and its expected value depends on the shoe composition and nothing else. The name is doing all the work: nothing is being insured, and there is no policy. Which is why the offer lands hardest when you have a hand worth protecting — exactly when it feels most reasonable, and exactly as bad as every other time.

Why 2 to 1 is not enough

A bet paying 2 to 1 breaks even when it wins one time in three. So insurance is profitable only when more than 33.33% of the unseen cards are tens. The Wizard of Odds states the condition in exactly those terms: on average, when the dealer has an ace up, the remaining cards will be 30.87% tens in a six-deck game, making insurance a bad bet, and it becomes a good bet only if that probability gets above 33.33%.

Three percentage points does not sound like much. It is the whole game. A deck is four-thirteenths tens — sixteen ten-value cards in fifty-two, counting every ten, jack, queen and king — or 30.8%. The casino pays 2 to 1 — a price that would be fair only if tens turned up a third of the time — on a shot that comes in a bit under 31 times in a hundred. That gap is where the house edge comes from, and because the bet pays 2 and loses 1, the edge works out to roughly three times it.

Work the single-deck case by hand, since it is short. The dealer's ace is face up, so 51 cards are unseen and 16 of them are tens: 31.37%. You win two units on 16 of 51 and lose one unit on the other 35. That is (32 minus 35) divided by 51, or -3/51 = -5.88%. The Wizard's single-deck strategy page publishes the same number rounded: 'Never take insurance or even money. The house edge on insurance is 5.9%, based on one deck.' The arithmetic above is mine; it reproduces his figure to the decimal, which is the point of showing it.

The six-deck version is the same shape with worse numbers, and the Wizard works it out card by card: 96 winning cards out of 311, 215 losers. Two times 96/311 is 61.74%, and 215/311 against you is -69.13%. Net: -7.4%.

Single deck: 16 tens of 51 unseen = 31.37% — house edge 5.9%.
Six decks: 96 tens of 311 unseen = 30.87% — house edge 7.4%.
Break-even for a 2-to-1 payout: 33.33%.

Fewer decks, better bet — and it still loses

Notice the direction of that difference. Insurance is genuinely less bad in a single deck than in a shoe, and the reason is visible in the fractions above: removing one card from the unseen pool — the dealer's ace, a non-ten — nudges the ten density from 30.77% up to 31.37% in a single deck, but only to 30.87% across six. One card is 2% of a deck and 0.3% of a shoe. Same mechanism that makes counting more powerful in fewer decks, and it generalizes: every card removed from a small pool moves the composition further than the same card removed from a big one.

But note the size of the effect. The best case in ordinary play is still a bet with a 5.9% edge against you. Single-deck insurance is not a good bet; it is a slightly less expensive bad one.

Even money is insurance in a costume

You have a blackjack, the dealer shows an ace, and before you can think you are offered even money — 1 to 1 right now, guaranteed, instead of risking a push. It is the same bet you just read about. Wikipedia says so plainly: taking even money is taking insurance with a blackjack, giving up a 3-to-2 payout for a 1-to-1 payout.

The equivalence is easy to check, and this worked example is mine. Bet $100, catch a blackjack, insure for the maximum $50. If the dealer has blackjack your hand pushes for nothing and insurance pays 2 to 1 on $50: you collect $100. If the dealer does not, your natural pays $150 and the insurance loses $50: you collect $100. Both branches land on exactly $100, which is why the casino saves itself the suspense and offers you the $100 up front. The cancellation depends entirely on the 3-to-2 payout — at a table paying 6 to 5 the same branches come out to $100 and $70, and 'even money' stops describing anything.

The reason to decline is the reason to decline insurance, and the Wizard of Odds puts it in units: 'the expected value of a blackjack when the dealer has an ace showing is 1.04 units, better than the 1.00 units you'll get by taking the even money. So unless you're a card counter and know the remaining deck to be 10-rich then always decline even money.' His video pins it at 103.88%, which you can rebuild — with your own ace and ten removed from a six-deck shoe, 214 of the 309 unseen cards are not tens, and 214/309 times 1.5 units is 1.0388.

He also names the real obstacle, which is not mathematical: 'The casinos here always offer even money in this situation and both dealer and other players will treat you like a fool if you decline it.' Four percent of your natural is the price of not being talked out of it.

The exception: when the count says yes

Everything above assumes you know nothing about the remaining cards. A counter does. Insurance is profitable the moment the hole card has better than a one-in-three chance of being a ten, and that is precisely what a high count describes — Wikipedia notes that card counting techniques can identify such situations, which is a dry way of saying this is the bet the whole enterprise was built for.

For Hi-Lo the threshold is published and unambiguous. The Wizard of Odds gives it in one line: 'The player should take insurance if the True Count is +3 or greater.'

Worth knowing, because most write-ups flatten it: the flat +3 is the multi-deck figure, not a universal constant. Michael Dalton's Encyclopedia of Blackjack gives the Hi-Lo thresholds by deck count — take insurance when the count per deck exceeds 1.5 for single-deck, 2.4 for double-deck, and 3 in multi-deck games. Same principle, three trigger points, and the reason is the one from two sections up: in fewer decks each card swings the composition further, so a smaller true count already means a ten-rich pool.

Below the threshold the answer does not soften into 'maybe'. At true count +2 in a six-deck shoe insurance is still a losing bet; it is just losing less than it was at zero. There is no partial credit for a count that is merely positive.

Hi-Lo, multi-deck shoe: take insurance once the true count reaches about +3. Otherwise decline.
Hi-Lo, double deck: the published threshold is +2.4.
Hi-Lo, single deck: the published threshold is +1.5.
Not counting, or counting but unsure of your true count: decline every time, including even money.

Why it is the most valuable index play

Don Schlesinger coined the term Illustrious 18 for the eighteen count-dependent departures from basic strategy that return the most. Insurance heads the list, ahead of 16 versus 10 and 15 versus 10, and it is not a narrow lead. Dalton's encyclopedia summarizes the finding for a typical four-deck shoe played with Hi-Lo: the insurance play alone is worth over 30% of all the gain, and the 'BIG 3' — insurance, 16 versus 10, 15 versus 10 — account for nearly 60%.

The Wizard of Odds' 'Why Card Counting Works' table, which he flags as initial results from an in-progress study, puts the same figure in a wider frame: of the total benefit of counting a six-deck shoe, standing accounts for 40%, insurance 34%, doubling 9%, blackjacks 7%, surrender 6%, splitting 4%. Worth noting that this is not a second, independent measurement — the Wizard states directly that his insurance probability was taken from Schlesinger's Illustrious 18 list, so it traces back to the same underlying work. Take it as one careful analysis, reported two ways: roughly a third of everything counting buys you rides on a bet the dealer only offers when an ace is showing.

There is a clean reason for the dominance. Every other index play asks a complicated question — given my two cards and the dealer's upcard, has the composition shifted enough to change what I do with this specific hand? Insurance asks one question, and it is the question the count already answers: are there a lot of tens left? And unlike a deviation that merely redirects a decision you were making anyway, insurance is an extra wager you get to place only when you know it is good.

One honest caveat on 'the count measures exactly this'. Hi-Lo's insurance correlation — which QFIT defines as the correlation between a system's point values and the value of cards in an insurance situation — is .76. High, not perfect. Hi-Opt I scores .85. Compare the tag sets and a likely reason appears: Hi-Lo tags the ace -1 alongside tens, while Hi-Opt I tags it 0, so only tens carry a negative value. An ace-rich, ten-poor shoe reads high in Hi-Lo and is bad for insurance. That inference is mine from the two tag lists, not a causal claim either source makes, but it is a useful reminder that +3 is a good rule rather than an oracle.

What insuring out of habit costs

Put a number on the bad version. The dealer's upcard is an ace about one hand in thirteen, or 7.69%. Insure for the maximum half-unit every time, at the six-deck edge of 7.4%, and the expected cost is 0.0769 x 0.5 x 0.074 = 0.28% of your original bet per hand. That derivation is mine, and it is deliberately rough — it treats the ace frequency as a flat 4 in 52 and applies an average shoe composition — but the order of magnitude is solid.

Now compare it to something. The Wizard's house-edge comparison puts blackjack under liberal Vegas rules at 0.28%. A habitual insurance player at a good table is paying roughly twice the house edge he thinks he is paying, and the second half is invisible because it arrives as a separate stack of chips that feels like caution rather than a wager.

That is the shape of the whole topic. Insurance is not a trap because it is complicated; it is a trap because declining feels like recklessness and taking it feels like prudence, and the arithmetic is the exact reverse. For a counter it flips — but only above a specific number, and only if the number is real.

Making the number real

The insurance index is the easiest deviation to memorize and the hardest to actually use, because it does not test your memory of the rule. It tests whether your true count is trustworthy under table conditions — a dealer waiting, other players deciding, a few seconds to divide a running count by the decks remaining and compare it to three. A true count you are unsure of is worse than none, because it will occasionally talk you into a 7.4% bet.

That conversion is the drillable part, and it is what the counting trainer here is for: it deals a shoe at the speed you choose, pauses at checkpoints, and grades your running count and, with the true-count conversion toggled on, your true count as well — in Hi-Lo from the start, with KO, Hi-Opt I and Omega II unlocking as you level up, at whatever deck count you set. The decision trainer covers count-dependent hand plays separately, things like 16 versus 10 and 12 versus 3. Neither drill deals you an insurance prompt today, so treat the +3 rule as something to memorize and the true count as the thing to train.

Insurance is one number away from being the best bet a counter makes and the worst one anybody else makes. The number is your true count — so make sure it is right.

Drill the true count free →

Share this post — the link unfurls with its own card.