A pair asks the strangest question in blackjack: would you rather play the hand you were dealt, or pay the price of a second bet to tear it in half and play two? Casual players answer with vibes, splitting "pretty" pairs and hoarding ugly ones. The chart answers with arithmetic, and its verdicts sort neatly into three drawers: always, never, and it depends on the dealer.
The two always splits
Aces, always. Kept together, two aces make soft 12, a hand going nowhere. Split, each ace starts a hand at 11, the best first card in the game, with the deck's densest value, the ten, sitting one draw away from 21. No dealer upcard changes this verdict. Casinos know how strong the split is, which is why they fence it: typically one card per ace, no re-hitting, and a ten on a split ace counts as a plain 21, not a 3:2 blackjack. Even fenced, the split is correct everywhere.
Eights, always. Two eights are hard 16, the game's worst total, the hand our hit-or-stand guide calls a choice between bad and worse. Splitting replaces one doomed 16 with two hands starting from 8, each of which can reach a respectable 18. Against a dealer 10 the split still loses on average, and it is still right: it loses less than any way of playing the 16, per the commonly published analyses. Splitting 8s is defense priced as aggression.
The two never splits
Tens, never. Twenty wins the overwhelming majority of hands it plays. Splitting it trades that near-certain winner for two hands that each need a strong draw to match what you already had, and the commonly published numbers score the trade as throwing money away against every upcard. (Card counters at very high counts have a famous exception; at every count a casual player will ever see, keep your 20.)
Fives, never. A pair of fives is a costume: underneath is hard 10, one of the best doubling hands in the game, as our doubling guide details. Split it and you own two 5s, each likely to grow into exactly the stiff totals you spend the rest of this game avoiding. Treat 5-5 as 10, double against 2 through 9, and never think of it as a pair again.
21 TrainerPairs are the heat map's favorite weak spot. Drill them with instant feedback, free on the App Store.
Get the app
The it-depends drawer
Everything else splits against weak dealer upcards and stays whole against strong ones, per the commonly published multi-deck DAS charts:
| Pair | Split against | Otherwise |
|---|---|---|
| 2s and 3s | 2 through 7 | Hit |
| 4s | 5 and 6 only | Hit |
| 6s | 2 through 6 | Hit |
| 7s | 2 through 7 | Hit |
| 9s | 2 through 6, 8 and 9 | Stand |
The nines row is the chart's best riddle. Why stand against a 7 but split against an 8? Because your 18 already beats the 17 that a dealer 7 most often makes, so you keep the made winner; against an 8 or 9 that advantage is gone, your 18 is an underdog or a coin flip, and two hands built from 9 outperform it. It is the dealer-script logic of basic strategy in miniature: you are never playing your cards, you are playing them against the ending the upcard predicts.
The rule that moves the borderline
That table assumed DAS: permission to double down after splitting. It matters because much of a split's value arrives when a split hand draws into a doubling spot, a 2 that becomes an 11, say. Take DAS away and the marginal splits lose their sweetener: 2s and 3s retreat to splitting only against 4 through 7, 4s stop splitting entirely, 6s against 3 through 6. Our DAS guide covers the rule's price in full; for chart purposes the moral is familiar from every rules discussion on this site: know which game your chart was computed for.
Splitting well, in practice
Pairs are rare enough that they never get natural practice volume: you see a splittable pair a few times an hour, which is exactly why pair cells dominate most players' error heat maps. They are also the most expensive category to fumble, since every split doubles the money in play. The fix is artificial volume: drills that deal you pairs on purpose, grade the decision instantly, and return to the ones you miss. Ten focused minutes on pairs, per the loop in our practice guide, outweighs months of table frequency, and it converts the three drawers, always, never, and it depends, from a list you recite into a reflex you own.
Frequently asked questions
Which pairs should you always split in blackjack?
Aces and eights, against every dealer upcard. A pair of aces is a mediocre soft 12 as one hand but two chances at 21 as two hands. A pair of eights is hard 16, the worst total in the game, while two hands starting from 8 are workable. Both splits are correct at every table, whatever the dealer shows.
Which pairs should you never split?
Tens and fives. A pair of tens is 20, a hand that wins on its own; breaking it trades a near-certain winner for two gambles. A pair of fives is really a hard 10, one of the best doubling hands in the game, and splitting it turns that asset into two weak hands likely to become stiffs.
Do you split 9s in blackjack?
Against 2 through 6 and against 8 and 9, yes; against 7, 10 and ace, stand, per the commonly published multi-deck charts. The odd shape has a logic: your 18 already beats the 17 a dealer 7 tends to make, so you keep it, but against an 8 or 9 your 18 is no longer a favorite and two hands of 9 do better.
Why split 8s against a 10 when it feels terrible?
Because the alternative is playing hard 16 against a 10, the worst spot in blackjack. Splitting does not turn the hand profitable; it loses less than hitting or standing a 16 does over the long run, per the commonly published analyses. Where late surrender is available, some rule sets prefer surrendering; otherwise the split is the least bad option.
What happens after you split aces?
Most casinos deal exactly one card to each ace, with no further hitting, no doubling, and often no resplitting if you draw another ace. A ten on a split ace also counts as an ordinary 21, not a blackjack, so it pays even money. The restrictions trim the value of the split, but even with them, splitting aces remains correct everywhere.