Splitting 8s vs an ace
A pair of eights against a dealer ace looks like the worst 16 in the game. It is not played as a 16. In almost every PlayCorrect rule mix, you split. The one well-known exception is a six- or eight-deck game where the dealer hits soft 17 and late surrender is offered: then the chart says surrender, otherwise split. Educational, 21+, not gambling advice.
| Rules | Play | Expected values (units) |
|---|---|---|
| 6-deck S17 DAS, no surrender (default) | Split | Split −0.364660, hit −0.513551, stand −0.663258 |
| 6-deck S17, late surrender on | Split (do not surrender) | Split −0.364660, surrender −0.500000 |
| 1- or 2-deck, S17 or H17, late surrender on or off (DAS on) | Split | 2-deck H17 DAS, no surrender: split −0.496112 |
| 6- or 8-deck H17, late surrender on | Surrender, otherwise split | 6-deck: surrender −0.500000, split −0.514859 |
| Same H17 late-surrender mix, 2-deck DAS | Split in our table | Split −0.496112, surrender −0.500000. Wizard of Odds prints Rp. See below. |
Why split even when both eights look weak
Standing on 16 vs ace is grim (default stand EV −0.6633). Hitting is better but still a large loss (−0.5136). Two hands that start with an eight, even against an ace, have a combined expected value of −0.3647 under default rules — still a loss, but a smaller one. That is the entire argument. It is not a claim that either split hand is likely to win.
With late surrender and S17, giving up half a unit (−0.500) is still worse than splitting (−0.3647). That is why six-deck S17 late-surrender charts keep the split, and why they matched Wizard of Odds 350/350 on 10 Nov 2026.
The H17 late-surrender exception
When the dealer hits soft 17, our six- and eight-deck split EV vs ace falls to about −0.515, which is worse than surrender at −0.500. The chart action becomes Rp: surrender if it is offered, otherwise split. Those six- and eight-deck H17 late-surrender tables also matched Wizard of Odds 350/350 on 10 Nov 2026.
On two-deck H17 DAS with late surrender, our engine still prefers the split by 0.00389 units (split −0.49611, surrender −0.50000). The Wizard of Odds chart prints surrender-else-split. We left our computed split in place and listed this as one of three one-cell disagreements on about accuracy. Non-ace split EVs use an independence approximation, so a gap this small is a fair place to be cautious.
Against a 10 rather than an ace, six-deck S17 still splits (split −0.4748, hit −0.5354). See also 16 vs 10, which is the non-pair hand.