Hardest Blackjack Hands: 12 Plays Ranked by What They Cost

The hardest blackjack hands are the ones where the natural play costs money, and the single costliest habit is taking insurance, at 0.28 units per 100 hands. Next come splitting a pair of tens against a 6 (0.18 units per 100 hands), hitting 11 against a 10 instead of doubling (0.08) and standing on 16 against a 10 instead of surrendering (0.07). This page ranks twelve common blackjack mistakes by what they cost over time, using the exact expected values from our engine.

The ranking multiplies two numbers: how much a wrong play loses each time it happens, and how often the situation comes up. That second number matters. The biggest single error here costs 0.49 units, but you meet the hand once in 2,271 rounds.

Units lost per 100 hands by always taking insurance, the costliest mistake

0.28

Units lost each time you hit A-A against a 6 instead of splitting

0.49

Hands that are a hard 16 against a dealer 10

1 in 59

Units per 100 hands if you got all twelve plays wrong every time

0.72

Source: Our study S5, src/data/studies/s5-blackjack-odds.json (mistakes[])

PlayCost each timeOnce inCost per 100 hands

1. Taking insurance 3.7 13 0.28
2. Splitting 10-10 vs 6 25.1 137.4 0.18
3. Hitting 11 vs 10 6.0 73.72 0.08
4. Standing on 16 vs 10 4.0 59.47 0.07

How we ranked them

All figures are for six decks, dealer stands on soft 17, double after split and late surrender, which is the common modern table. A “unit” is your original bet. Take a ten-dollar bet: 0.28 units per 100 hands is $2.84 over $1,000 wagered.

For each play we took the engine’s expected value (EV) of every legal action in that spot. The cost of a mistake is the best action’s EV minus the EV of the mistaken one. Multiplying by how often the spot arises gives the cost per 100 hands. Those frequencies are computed by counting the exact two-card hands and the dealer upcard, then removing dealer naturals. They are not survey data about how often players really err, so the final column assumes you make that exact mistake every time it appears. Real players are right some of the time, so read the column as a ceiling.

Rank	The mistake	Best play	Cost each time (units per 100 bet)	Comes up once in (hands)	Cost per 100 hands (units)
1	Taking insurance	Decline	3.70	13.0	0.284
2	Splitting 10-10 vs 6	Stand	25.08	137.4	0.183
3	Hitting 11 vs 10	Double	6.00	73.7	0.081
4	Standing on 16 vs 10	Surrender	4.01	59.5	0.067
5	Standing on 16 vs 7	Hit	6.80	219.4	0.031
6	Not splitting A-A vs 6	Split	49.16	2,271.0	0.022
7	Standing on 12 vs 3	Hit	1.83	156.4	0.012
8	Not splitting 8-8 vs 10	Split	6.11	615.4	0.010
9	Hitting 13 vs 2	Stand	1.52	155.4	0.010
10	Hitting 10 vs 9	Double	3.15	362.7	0.009
11	Standing on A-7 vs 9	Hit	8.34	1,088.0	0.008
12	Standing on A-7 vs 3	Double	2.77	1,088.0	0.003

Cost each time is the best action's EV minus the mistaken action's EV, from chart-cell EVs averaged over two-card compositions. Frequency counts only two-card starting hands. Taking insurance is costed on a half-bet insurance stake.

Source: Our study S5 (mistakes[])

Source: Our study S5 (mistakes[].costPer100HandsIfAlwaysMadeUnits)

See the numbers as a table

Cost per 100 hands of making each mistake every time it comes up

	Units lost per 100 hands
Taking insurance	0.284
Splitting 10-10 vs 6	0.182
Hitting 11 vs 10	0.081
Standing on 16 vs 10	0.067
Standing on 16 vs 7	0.031
Not splitting A-A vs 6	0.022
Standing on 12 vs 3	0.012
Not splitting 8-8 vs 10	0.010
Hitting 13 vs 2	0.010
Hitting 10 vs 9	0.009
Standing on A-7 vs 9	0.008
Standing on A-7 vs 3	0.003

The four that cost the most

1. Taking insurance: 0.2844 units per 100 hands

Insurance is a separate bet that pays 2 to 1 if the dealer’s hole card is a ten. It only breaks even if that happens one time in three, 33.33%. Against an ace, the chance is 30.87%, so the bet loses 0.0740 per unit insured and costs 0.0370 of your original bet at the usual half-bet stake. Dealer aces show once in 13 hands, so this is the most frequent decision on the list. It is a side bet, not protection for your hand. The guide to how table rules move the edge shows where insurance sits next to them.

2. Splitting 10-10 against a 6: 0.1825 units per 100 hands

Twenty is already a strong total. Against a 6 the dealer busts 42.28% of the time, and standing is worth 0.7028 units. Splitting turns one hand likely to win into two hands that start from a 10, worth 0.4520 in total. It is still a profitable play, which is exactly why it tempts people, but it gives up 0.2508 units every time. The spot arises once in 137.4 hands, far more often than the other pair mistakes here, so it ranks second.

3. Hitting 11 against a 10 instead of doubling: 0.0815 units per 100 hands

Both options win. Hitting is worth 0.1193 and doubling 0.1793, so hitting loses 0.0600 per hand, from a spot that appears once in 73.72 hands. The engine explains it as a doubling spot where you are the favourite to finish with the better hand, so more money out earns more. The mistake feels safe because the dealer’s 10 looks scary, but the 10 is also a card that makes your 11 into 21.

4. Standing on 16 against a 10 instead of surrendering: 0.0675 units per 100 hands

Surrendering costs half your bet, so its EV is exactly −0.5. Standing is −0.5401, hitting −0.5346. All three are losing plays, and surrendering loses least. Standing costs 0.0401 more than surrendering, and 0.0056 more than hitting, at a rate of once in 59.47 hands. If the table does not offer late surrender, hit. Surrendering depends on the rules, so the ranking above is specific to tables that allow it.

The expensive hand you rarely see

Not splitting a pair of aces against a 6 has the largest cost of all per hand: 0.4916 units, the gap between splitting (0.6800) and hitting a soft 12 (0.1885). Aces split into two hands that each start with an ace, which is the best starting card in the game. But a pair of aces against a 6 comes up once in 2,271 hands. So its cost per 100 hands is only 0.0217, and it ranks sixth.

Source: Our study S5 (mistakes[].costPerHandUnits)

See the numbers as a table

Cost each time you make the mistake (units per 100 bet)

	Units lost each time
Not splitting A-A vs 6	49.16
Splitting 10-10 vs 6	25.08
Standing on A-7 vs 9	8.34
Standing on 16 vs 7	6.80
Not splitting 8-8 vs 10	6.11
Hitting 11 vs 10	6.00
Standing on 16 vs 10	4.01
Taking insurance	3.70
Hitting 10 vs 9	3.15
Standing on A-7 vs 3	2.77
Standing on 12 vs 3	1.83
Hitting 13 vs 2	1.52

Compare it with the ranking. Three of the five plays that cost most each time (A-A against a 6, A-7 against a 9, 8-8 against a 10) come up once in 600 hands or more, and none of the three most frequent spots (insurance, 16 against a 10, 11 against a 10) is in the top five for cost each time. That is why a fixed list of “never do this” rules misleads: weight by how often you actually see the hand.

Source: Our study S5 (mistakes[].frequencyPerHand)

See the numbers as a table

How often each spot comes up (times per 1,000 hands)

	Times per 1,000 hands
Taking insurance	76.9
Standing on 16 vs 10	16.8
Hitting 11 vs 10	13.6
Splitting 10-10 vs 6	7.3
Hitting 13 vs 2	6.4
Standing on 12 vs 3	6.4
Standing on 16 vs 7	4.6
Hitting 10 vs 9	2.8
Not splitting 8-8 vs 10	1.6
Standing on A-7 vs 9	0.9
Standing on A-7 vs 3	0.9
Not splitting A-A vs 6	0.4

The middle of the list

The next six plays each cost between 0.0087 and 0.0310 units per 100 hands.

• Standing on 16 against a 7 (0.0310). The dealer busts only 26.19% of the time with a 7 showing, so standing rarely wins. Hitting is −0.4092 and standing −0.4773, a gap of 0.0680 units, once in 219.4 hands. A 16 against a 7 is a poor hand either way, but hitting is better.
• Not splitting A-A against a 6 (0.0217), covered above.
• Standing on 12 against a 3 (0.0117). Against a 3 the dealer busts 37.42%. Hitting a 12 risks a 31% bust, which is why standing feels safe. The engine says hitting is worth −0.2338 against −0.2521, gaining 0.0183 per hand.
• Not splitting 8-8 against a 10 (0.0099). Hard 16 is the worst total in the game, so two hands that start from 8 do better. Splitting is −0.4757 and standing −0.5369, a gain of 0.0611, but it asks you to put a second bet out against a dealer 10.
• Hitting 13 against a 2 (0.0098). This is a mistake in the opposite direction: stand. Standing is −0.2931 and hitting −0.3083, a gap of only 0.0152.
• Hitting 10 against a 9 (0.0087). Doubling is 0.1495 and hitting 0.1180, a gap of 0.0315. The spot arises once in 362.7 hands.

The soft 18 pair

Two last plays involve ace-seven and move in opposite directions. Standing on A-7 against a 9 costs 0.0834 each time (0.0077 per 100 hands): hitting is −0.0992 against −0.1826 for standing. Soft 18 cannot bust with one card, so the draw is free, and a 9 is a strong dealer card. Standing on A-7 against a 3 costs 0.0278 (0.0026 per 100 hands) because doubling is worth 0.1789 and standing 0.1511. The soft 18 is the one hand total where three different plays are right depending on the upcard, which is why it gets misplayed. The complete recommended play for every hand is on one page.

Hand vs dealer	Stand	Hit	Double	Split	Surrender
10-10 vs 6	0.7028	−0.8522	−1.7045	0.4520	−0.5000
A-A vs 6	−0.1457	0.1885	0.1900	0.6800	−0.5000
11 vs 10	−0.5398	0.1193	0.1793	-	−0.5000
16 vs 10	−0.5401	−0.5346	−1.0691	-	−0.5000
16 vs 7	−0.4773	−0.4092	−0.8184	-	−0.5000

A dash means the action is not legal. 16 vs 10 is a surrender spot; the others favour split, double or stand as listed in the ranking table.

Source: Our study S5 (mistakes[].ev)

What these mistakes cost over a session

Add up all twelve, assuming a player gets every one of them wrong every time, and the cost is 0.72 units per 100 hands. The house edge of the table itself is 0.33 units per 100 hands, so an always-wrong player pays more than twice the edge on top of it. Without insurance the total is 0.43 units.

At $10 a hand and 70 hands an hour (our pace assumption), the cost of always being wrong on all twelve is about $5.03 an hour on top of the table’s $2.32. Nobody makes every one of them every time, so treat these as ceilings. The point of the list is where practising helps most: insurance, the 10-10 split, doubling 11 and surrendering 16.

How we checked the numbers against published figures

Wizard of Odds publishes composition-dependent EVs for six decks and a dealer who stands on soft 17. We compared each of the twelve plays with the matching rows. Our engine uses chart-cell averages over two-card compositions, so small differences are expected.

Play	Engine EV	Wizard of Odds EV (range)	Verdict
10-10 vs 6, stand	0.7028	0.7028	Equal
10-10 vs 6, split	0.4520	0.4449	Engine 0.007 higher; cost 0.2508 vs 0.2579 from their figures
A-A vs 6, split	0.6800	0.6824	Within 0.003
11 vs 10, double	0.1793	0.1739 to 0.1785	Within 0.001
11 vs 10, hit	0.1193	0.1170 to 0.1186	Within 0.001
16 vs 10, stand	−0.5401	-0.5368 to -0.5410	Inside the range
16 vs 7, hit	−0.4092	-0.4084 to -0.4086	Within 0.001
8-8 vs 10, split	−0.4757	-0.4754	Within 0.001
A-7 vs 9, stand	−0.1826	-0.1826	Equal
12 vs 3, hit	−0.2338	-0.2314 to -0.2370	Inside the range
13 vs 2, hit	−0.3083	-0.3054 to -0.3117	Inside the range

Wizard of Odds lists each two-card composition separately, so a range is shown for hands such as 10-6 and 9-7. Hand probabilities in their table also match our frequencies: 10-10 vs 6 is 0.007277, hard 16 vs 10 is 0.016814 for 10-6 plus 9-7.

Source: Wizard of Odds, blackjack appendix 9 (6 decks, S17), fetched 9 October 2026; our study S5

The only meaningful difference is the 10-10 split, where our engine is 0.007 higher than theirs. We kept it: the correct play and the ranking do not change, and the cost is within 3% of the figure from their table. Everything else lies inside or within 0.003 of the published range. The engine also gives the same dealer-natural-free frequencies as their probability column.

What to do with this

Aim your practice at frequency first. Decline insurance and even money, stand on 20, double 11 against a 10, and learn the one surrender spot, hard 16 against a 10. After that, the rare plays such as A-A against a 6 are worth learning only because they are easy to memorise. The soft-17 variant of the strategy changes a few of these doubles and surrenders, and the one-deck version changes others, so check which table you are on.

What we could not verify, and our assumptions

• Frequency is computed, not observed. We enumerate the cards. We did not survey real players, so we cannot say how often people actually make each mistake. The cost-per-100-hands column assumes you make the mistake every time.
• Two-card hands only. A hard 16 reached after hitting 5-4-7, or a 13 made of three cards, is not counted, so frequencies for hard totals understate real exposure. Costs for those compositions differ slightly because removed cards shift the odds.
• Chart-cell averages. The engine averages over the two-card compositions that make each total and does not remove the player’s own cards from the shoe. Wizard of Odds uses exact compositions. Differences stayed within 0.003 except the 10-10 split (0.007).
• Insurance was not checked against a fetched external page. Its cost follows from the 30.87% chance of a ten under an ace and the 2-to-1 payout, and matches the 7.40% figure used elsewhere on this site. It ignores your own cards.
• Twelve plays, not more. We kept the twelve in our study. We did not rank other mistakes such as never doubling soft hands, incorrect splits of 4-4 or 6-6, or mistakes arising only with three or more cards.
• Rule dependence. Late surrender is assumed. Without it, standing on 16 against a 10 costs only 0.0056 units per hand and drops down the ranking. The ranking also changes with other rules, which we did not rank.
• Explain text. The engine’s plain-English explanation is written for the best play. We used it to describe why, but its “worth X more than Y” comparison is against the natural alternative, which is not always the mistake, so every cost on this page uses the study’s own mistake figures.
• Dollar figures use $10 hands and 70 hands an hour, our assumptions, not measured table pace.