Blackjack: When to Split Pairs — The Complete Splitting Strategy
Aces and eights always, fives and tens never — and the rest? We walk through every possible pair, explain the math, and show where the most common splitting mistake costs 0.8% house edge.
Reviewed by the Casinokeller editorial team · Editorial policy
Splitting is the only blackjack decision that doubles your stake AND turns one hand into two new ones. Used correctly it is among the most valuable plays; used wrong it costs more than any other mistake. The rule 'aces and eights always, fives and tens never' covers about 80% of cases. The rest we cover here.
What splitting does mathematically: You trade a hand worth X for two new hands each starting at X/2. That makes sense only when (a) your current hand is bad (16 is weaker than two starting 8s) OR (b) your starting card is strong enough that two tries beat one (A,A = two starts of 11). In any other case, you give up money.
A,A — always split: A soft 12 is mediocre; two hands starting at 11 are extremely strong. Most casinos give only one extra card on each split ace — even so the expected value is far higher. Never stand, never play as a 12.
8,8 — always split: 16 is the worst hand in blackjack — you lose to any dealer 7+. Two new hands starting at 8 are only 'so-so' but better than a guaranteed loser. Split even against dealer 10 or ace — the alternative is worse. Surrender is the only exception there, if allowed.
10,10 — never split: 20 wins about 85% of the time long-term. Two new hands starting at 10 have lower expected value — you would be giving up profit, not making more. The 'split 10s vs. 5/6' myth applies only to card counters in extreme positive counts. For normal players: stand.
5,5 — never split: A pair of 5s totals 10 — one of the best doubling hands. Splitting turns a top hand into two mediocre ones. Play as a 10: double vs. 2–9, hit vs. 10/A.
9,9 — split vs. 2–9 except 7: Sounds odd, is precise. An 18 loses long-term to dealer 9, 10, A. Splitting gives two hands starting at 9 — better odds. Against dealer 7, 18 is strong enough (dealer often lands on 17), so stand. Against 10/A, splitting is too risky.
7,7 — split vs. 2–7: 14 is loser territory. Split against weak dealer cards (2–6) and also vs. 7 (dealer often makes 17 and your 17 after hit wouldn't win). Vs. 8–A hit — splitting would create two losing hands.
6,6 — split vs. 2–6: Same principle. 12 is marginal; splitting against bust cards (2–6) is more profitable. Vs. 7+ hit. If DAS is allowed, splitting also vs. 7 is marginally positive.
4,4 — mostly do NOT split: An 8 is a fine hit hand; two 4s are weak starts. Split only if DAS is allowed AND dealer shows 5 or 6 — then the doubling upside makes it worthwhile. Otherwise hit.
3,3 and 2,2 — split vs. 2–7: A 4 or 6 is just a hit hand with low value. Two 2s/3s vs. weak dealer cards give two profitable starts. Vs. 8+ hit.
Most common splitting mistakes: Splitting 4,4 ('they're small cards anyway') costs about 0.8% EV without DAS. Splitting 10,10 ('two shots at 20!') costs about 1.2%. Splitting 5,5 costs about 0.4% in forfeited doubling upside.
Key house rule — Double After Split (DAS): When DAS is allowed, more splits become profitable (especially 2,2, 3,3, 4,4, 6,6). When not, some splits drop below the value of hitting — don't split them. Listed in the table rules or printed on the felt.
Related: 'Blackjack basic strategy — the complete chart', 'Blackjack house edge with basic strategy', 'Is blackjack actually beatable?'. Tool: bankroll simulator to see the effect of correct splits across 1,000 hands.
Bottom line: Splitting is powerful but not a cure-all. The five core rules (A,A always, 8,8 always, 5,5 never, 10,10 never, 9,9 almost always except 7/10/A) cover most of the value. The rest follows 'split vs. weak dealer cards, hit vs. strong'.