♠️
The short version: splitting is correct in 52 of the 100 pair-versus-upcard situations. Four rules are absolute and never change: always split A,A and 8,8, never split 10,10 or 5,5. Everything else is a margin, and 8 of the hundred are decided by less than 0.03 per unit — close enough that the table rules matter more than the choice.

Every Pair, Every Upcard

Your pair down the left, the dealer’s upcard across the top. Green means split. Multi-deck, dealer stands on soft 17, double after split allowed, one split only, no surrender — every number below is computed for exactly those rules.

Your pair 2 3 4 5 6 7 8 9 10 A
A,A SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT
10,10 stand stand stand stand stand stand stand stand stand stand
9,9 SPLIT SPLIT SPLIT SPLIT SPLIT stand SPLIT SPLIT stand stand
8,8 SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT
7,7 SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT hit hit hit hit
6,6 SPLIT SPLIT SPLIT SPLIT SPLIT hit hit hit hit hit
5,5 double double double double double double double double hit hit
4,4 hit hit hit SPLIT SPLIT hit hit hit hit hit
3,3 SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT hit hit hit hit
2,2 SPLIT SPLIT SPLIT SPLIT SPLIT SPLIT hit hit hit hit

Computed from 340 solved hands · expected values per unit staked.

The Four Rules That Never Change

Most of basic strategy is a close call. These four are not, and they are the only pair decisions you can make without looking anything up.

Always split A,A

Two aces played as one hand is a soft 12, the worst total in the game. Split them and each ace starts a hand that makes 21 with any ten — and there are sixteen tens in a deck. Splitting is correct against all ten upcards, by a wide margin every time. Against a 6 the gap is 0.4814 per unit.

Always split 8,8

Sixteen is the worst hand in blackjack, and this is the one time you are allowed to stop holding it. Two eights become two hands starting at 8 — not good, but not 16. This holds even against a 10, where both halves are still losers: you are choosing the smaller loss, not a win. The most expensive mistake on the whole grid is failing to split these against a 7, which costs 0.6263 per unit.

Never split 10,10

You are holding 20. It wins outright most of the time. Splitting trades a near-certain winner for two hands starting at 10, and the arithmetic is brutal: against a 9 it costs 0.5253 per unit, the second most expensive error available on this chart. Players do it anyway, because two hands feel like twice the chance. They are twice the exposure on a worse position.

Never split 5,5

Two fives is a hard 10 — one of the strongest doubling hands there is. Splitting throws that away to start two hands at 5, the worst starting card. Treat 5,5 as a 10 and double it wherever doubling a 10 is correct.

The Eight Closest Calls

These are the hands people argue about, and the arguing is reasonable: the two options are nearly indistinguishable. On any of these, your table’s rules on doubling after a split will move the answer further than the choice itself does.

HandCorrect playRunner-upGap per unit
3,3 vs 2SPLIThit0.0026
4,4 vs 5SPLIThit0.0095
3,3 vs 8hitSPLIT0.0124
2,2 vs 8hitSPLIT0.0148
7,7 vs 8hitSPLIT0.0171
2,2 vs 2SPLIThit0.0260
5,5 vs 9doublehit0.0278
9,9 vs 7standSPLIT0.0295

The Six Most Expensive Mistakes

The other end of the same scale. Each of these costs more than 0.47 per unit staked — roughly half your bet, every time you get it wrong.

HandCorrect playCommon errorCost per unit
8,8 vs 7SPLIThit0.6263
10,10 vs 9standSPLIT0.5253
10,10 vs 10standSPLIT0.5039
10,10 vs AstandSPLIT0.4926
A,A vs 6SPLIThit0.4814
8,8 vs 6SPLITstand0.4792

Notice that four of the six involve the two absolute rules. That is not a coincidence: the rules are absolute precisely because the margins behind them are enormous. If you learn nothing else about splitting, learn those four and you have captured most of the available value.

Why the Margin Matters More Than the Answer

A strategy chart gives you one letter per cell and tells you nothing about how much that letter is worth. The margins above show why that is a loss of information.

Getting 3,3 against a 2 wrong costs 0.0026 per unit. Getting 8,8 against a 7 wrong costs 0.6263 — two hundred and forty times as much. Both are a single letter on a chart. Treating them as equally important is how players end up memorising the whole grid and still leaking money on the four hands that actually matter.

The practical consequence: learn the four absolutes cold, learn the six expensive errors, and let the close calls be close. On the eight hands at the top of that first table, playing the runner-up costs you less than a thirtieth of a unit — less than the house takes from you on a single spin of a roulette wheel.

Reality check: correct splitting does not make blackjack a winning game. It reduces the house edge to roughly 0.5% with the rest of basic strategy, which is the cheapest game on most casino floors — and still a negative-expectation one. The numbers above are computed for the rule set named at the top; a table with different rules has different answers, and the rules are worth more than the strategy.

Play with the math on your side

SlotDrop's free tools show you the real house edge, EV and variance behind every bet — before you place it.

Open the Black Box →

Reality check: no strategy on this page turns a negative-expectation game into a positive one. Strategy reduces how much the house takes and how fast your bankroll disappears — it does not guarantee profit. Only ever bet money you can afford to lose entirely.