Sorted by feel, blackjack hands split easily into good and bad. A 19 is good. A 16 is bad. A 17 is fine, probably. The real numbers exist and they are exact. After the dealer has checked and does not have blackjack, 14 of the 34 starting hands make money on average and 20 lose, and the 17 that feels fine sits closer in value to hard 12 than to hard 18.
Expected value is the average result of the round, in percent of the starting bet, over a long run of the same hand; our primer on expected value in blackjack builds it from scratch. Two independent sources put a number on every starting hand.
Two sources, one answer
The first source is Wizard of Odds blackjack appendix 9: the exact expected value of every two-card starting hand against every dealer upcard, computed combinatorially, six decks, dealer stands on soft 17, resplitting to four hands except aces, one card to split aces, best play for the exact cards held, from a full shoe, for rounds where the dealer has checked and does not have blackjack. With double after split, the expected return of that whole game is -0.4026 percent. We grouped the table by starting hand, weighting each two-card combination by its probability (our arithmetic on their table), and the grouped values add back to exactly -0.4026.
The second is our own simulation: about 62 million rounds through a six deck shoe with a cut card at 75 percent, basic strategy from the app's 4 to 8 deck chart, the same rules including double after split, and no surrender, the engine behind our piece on what share of rounds you win, push and lose. On rounds where the dealer has no blackjack, it agrees with the grouped Wizard of Odds numbers to within half a point. It also gives each hand's value counting dealer blackjacks, which the exact table leaves out.
All figures are in percent of the starting bet. The first two columns cover rounds after the peek, meaning the dealer has checked and has no blackjack; the third counts all rounds, dealer blackjacks included.
Every starting hand, ranked
Three tables cover the whole game: the natural and ten pairs, the eight soft hands, and the fifteen hard totals built from non-pair cards, so hard 20 appears only as 10,10 and hard 16 here means a 16 from other cards. Count the rows: 14 hands in the money, 20 out.
| Hand | Wizard of Odds | Simulation | Simulation, all rounds |
|---|---|---|---|
| Natural | +150.00 | +150.00 | +143.15 |
| 10,10 | +65.91 | +65.92 | +58.12 |
| A,A | +38.45 | +38.11 | +32.06 |
| 5,5 | +27.48 | +27.24 | +21.12 |
| 9,9 | +12.11 | +12.58 | +7.09 |
| 4,4 | -7.47 | -7.94 | -12.34 |
| 8,8 | -7.58 | -7.31 | -11.74 |
| 2,2 | -9.86 | -10.31 | -14.61 |
| 3,3 | -14.52 | -14.41 | -18.55 |
| 6,6 | -20.61 | -20.46 | -24.28 |
| 7,7 | -21.01 | -21.14 | -24.90 |
| Hand | Wizard of Odds | Simulation | Simulation, all rounds |
|---|---|---|---|
| Soft 20 (A,9) | +65.80 | +65.90 | +58.30 |
| Soft 19 | +32.85 | +32.85 | +26.75 |
| Soft 18 | +8.45 | +8.74 | +3.71 |
| Soft 13 (A,2) | +2.18 | +2.00 | -2.64 |
| Soft 14 | -0.61 | -0.65 | -5.24 |
| Soft 17 | -3.09 | -3.08 | -7.52 |
| Soft 15 | -3.50 | -3.66 | -8.11 |
| Soft 16 | -6.23 | -6.21 | -10.49 |
| Hand | Wizard of Odds | Simulation | Simulation, all rounds |
|---|---|---|---|
| Hard 11 | +37.86 | +37.94 | +31.29 |
| Hard 19 | +33.08 | +33.00 | +26.68 |
| Hard 10 | +27.18 | +27.16 | +21.04 |
| Hard 9 | +4.91 | +4.72 | -0.30 |
| Hard 18 | +4.32 | +4.28 | -0.67 |
| Hard 8 | -8.23 | -8.38 | -12.78 |
| Hard 7 | -17.48 | -17.48 | -21.45 |
| Hard 5 | -17.96 | -18.02 | -21.96 |
| Hard 6 | -20.16 | -20.31 | -24.12 |
| Hard 17 | -25.60 | -25.69 | -29.22 |
| Hard 12 | -28.14 | -28.17 | -31.60 |
| Hard 13 | -31.42 | -31.46 | -34.73 |
| Hard 14 | -34.01 | -34.02 | -37.16 |
| Hard 15 | -36.48 | -36.47 | -39.51 |
| Hard 16 | -38.55 | -38.54 | -41.46 |
The hands that feel safer than they are
Hard 17, a total that feels safe to stand on, averages -25.60 percent of the starting bet after the peek; only the non-pair stiffs, hard 12 through 16, lose more. Hard 18 does make money, at +4.32, but barely. Standing on 17 remains correct: the table reports what the hand is worth when played correctly, not a better way to play it.
The worst starting hand in the game is hard 16 at -38.55, with hard 15 close behind at -36.48; every non-pair total from 12 through 16 loses more than 28 percent of the bet on average. Against that sits the pair of 8s, which the chart splits and which is worth -7.58: two live hands starting from 8 instead of one hopeless 16. That gap is consistent with the published case to always split aces and 8s. One caveat: 8,8 and a hard 16 from other cards are not identical hands, because the compositions differ and the split puts a second bet into play.
21 TrainerFree on iPhone and Android. Instant right or wrong feedback with a clear explanation on every hand.
Get the app
What the dealer's blackjack changes
The exact table describes rounds after the peek. But in 4.75 percent of all rounds the dealer turns over a blackjack and the hand ends before it starts, and those rounds drag every starting hand down. Counting them, three hands flip from winners to losers in our simulation: hard 18 falls from +4.28 to -0.67, hard 9 from +4.72 to -0.30, soft 13 from +2.00 to -2.64.
The natural's +150.00 counts only rounds where the dealer does not have blackjack; counting the rounds where both sides are dealt one, the simulation puts it at +143.15, because the dealer also holds a natural in about 4.6 percent of the rounds where you are dealt one (one minus 143.15 divided by 150 is 4.57 percent), and those are pushes. The arithmetic behind that double natural frequency is in our piece on how often blackjack hands push.
Where the house edge comes from
Weight each starting hand by how often the deal produces it, the subject of our starting hand frequencies piece, and the whole game reassembles from these numbers. The table groups the Wizard of Odds values into families and their contribution per 100 rounds. Hand rows leave out dealer blackjacks, which get their own row, and 6,6, 7,7 and 8,8 sit under other pairs, so shares run below the starting hand frequencies. Rows are rounded.
| Hand family | Share of rounds | Average value | Units per 100 rounds |
|---|---|---|---|
| Natural | 4.532 | +150.00 | +6.798 |
| 10,10 | 8.956 | +65.91 | +5.903 |
| Hard 9 to 11 | 11.301 | +24.77 | +2.799 |
| Hard 18 and 19 | 9.046 | +18.70 | +1.692 |
| Soft 13 to 20 | 9.060 | +11.98 | +1.085 |
| Other pairs | 4.876 | -0.32 | -0.015 |
| Hard 5 to 8 | 6.781 | -14.92 | -1.012 |
| Hard 17 | 5.653 | -25.60 | -1.447 |
| Hard 12 to 16 | 35.046 | -33.31 | -11.674 |
| Dealer blackjack | 4.749 | -95.44 | -4.532 |
| Total | -0.4026 |
The winning families pour in 18.28 units per 100 rounds and the losing families take out 18.68; the difference is the 0.40 percent house edge. Hard 12 through 16 is 35.046 percent of all rounds and costs 11.674 units per 100, the largest entry in either direction, more than the dealer's blackjacks plus every other losing family combined (4.532 + 1.447 + 1.012 + 0.015 = 7.006). The naturals and the tens bring in 6.798 + 5.903 = 12.70 together, and the stiffs alone give back about 92 percent of that. The dealer blackjack row averages -95.44 rather than -100 because a player natural against it pushes, and it alone costs 4.532 units per 100.
Averages hide the upcard
Every value above averages all ten upcards; the spread underneath is enormous. A few upcard values from our grouping of the exact table, after the peek: hard 17 against a 6 is +0.8, the only upcard where our simulation also shows it above zero at +0.6, but against a 10 it is -41.8 and against an ace -47.6. Hard 18 against an 8 is +10.4, yet against a 9 it is -18.5 and against a 10 -17.5. Hard 11 ranges from +67.7 against a 6 to +14.7 against an ace. Hard 12 is worth -15.4 against a 6 but -25.3 against a 2. Hard 16 against a 10 sits at -53.5, a hand with its own full treatment. Our upcard series summary sets all ten chart columns side by side, with a link to each upcard's full treatment.
Reading the numbers honestly
Three caveats. The values assume correct basic strategy: composition dependent play in the exact table, the app's total dependent chart in the simulation. Other rules shift the values, and charts should be read for the rules actually in play. And these are long-run averages: a single hand is still won, pushed or lost, and nothing here is betting advice.
The values hold only for correct play, and the app gives instant right or wrong feedback with an explanation on every decision.
See the chart for your rulesFree on iPhone and Android. Color-coded charts that redraw automatically for the rules you select.
Download
Frequently asked questions
Which blackjack starting hands actually make money?
Fourteen of the 34 starting hands are positive after the dealer checks and has no blackjack: the natural, 10,10, soft 20, A,A, hard 11, hard 19, soft 19, 5,5, hard 10, 9,9, soft 18, hard 9, hard 18 and soft 13. The natural leads at +150.00 percent. The other 20 hands lose on average, worst of all hard 16 at -38.55 percent.
Is 17 a good hand in blackjack?
No. Hard 17 averages -25.60 percent of the starting bet after the dealer checks for blackjack, and only the non-pair stiffs, hard 12 through 16, lose more. Standing is still correct: the tables show what the hand is worth when played correctly, not a better way to play it.
What is the worst starting hand in blackjack?
Hard 16, at -38.55 percent of the starting bet on average after the peek. Hard 15 is close behind at -36.48, and every non-pair total from 12 through 16 loses more than 28 percent of the bet. A pair of 8s, which the chart splits, is worth only -7.58, a gap consistent with the case for always splitting 8s.
Why is a blackjack worth less than 150 percent?
A natural pays 3 to 2, but in about 4.6 percent of the rounds where you are dealt one the dealer also has a blackjack, and those hands push. Counting those rounds, our simulation values the average natural at +143.15 percent, and one minus 143.15 divided by 150 gives the 4.57 percent push share. The exact table counts only rounds where the dealer has no blackjack.
Do these hand values change with different rules?
Yes. The figures assume six decks, dealer stands on soft 17, double after split, resplitting to four hands except aces, one card to split aces. Different rules shift the values, so charts should be read for the exact rules in play. Misplayed hands are worth less than these numbers, which assume correct basic strategy.