Exact-card plays

Most charts are total-dependent: every hard 12 versus a dealer 4 gets the same play, whether the 12 is 10,2 or 5,7. A few holdings play the other way, and a few chart-correct plays still surprise people who already know basic strategy. The calculator marks those with a small sparkle and a one-line why, and links here.

These marks use only your cards, the dealer’s upcard, and the rule chips — not the rest of the shoe, and not a count. The list below is curated from published basic-strategy research (chiefly Wizard of Odds composition-dependent and strategy pages) and cross-checked against our already-computed tables. It is not every engine exception, and it is not a counting guide.

Hands that aren’t what they seem

Each item cites a source. Where our table was already checked against the Wizard of Odds hand calculator on 10 Nov 2026, that is noted. Sparkle marks appear only when the displayed play agrees with the fact.

• Even money / insurance. Decline insurance, and decline even money even when you already have a blackjack — they are the same side bet. Source: Wizard of Odds, 4–8 deck basic strategy (“Never take insurance or even money”), also the Wizard FAQ.
• Multi-card 16 vs 10: stand. The two-card chart hits 16 vs 10 (or surrenders, if allowed). With three or more cards, standing is usually better. Source: Wizard 2–8 deck composition-dependent summary and Ask the Wizard #277. We mark a holding only when our table actually stands; some three-card 16s still hit.
• 10,2 vs 4: hit. The chart stands on 12 vs 4; 10,2 is the composition that prefers a hit. Source: Wizard glossary and the one-deck CD exceptions. Our 1-, 2-, and 6-deck tables agree (verified 10 Nov 2026). Not an exception in our 8-deck table.
• 8,4 or 7,5 vs 3, single deck: stand. A single-deck 12 vs 3 is usually a hit, but these two holdings prefer to stand. Source: Wizard one-deck CD (H17) two-card exceptions. Our one-deck S17 table agrees (verified 10 Nov 2026).
• 10,3 vs 2, single deck: hit. The chart stands on 13 vs 2; a single-deck 10,3 prefers a hit. Source: the introduction to Wizard one-deck S17 CD strategy. Our one-deck table agrees (engine-computed; not yet a hand-calculator spot-check).
• Multi-card soft 18 vs ace: stand. Two-card A,7 vs ace is a hit in a shoe game; some later soft 18s prefer to stand. Source: the same 2–8 deck summary (3+ cards in two-deck S17; 4+ in four- or six-deck S17). We mark only holdings our table stands.
• Soft 18 vs 2 depends on S17 vs H17. Stand when the dealer stands on soft 17; double when the dealer hits soft 17 (2-, 6-, and 8-deck tables). Source: Wizard 4–8 deck chart. Our one-deck H17 table still stands, so that case is not marked as a double.
• Soft 18 vs 9, 10, or ace: hit in the usual shoe game, even though 18 looks made. Source: the same 4–8 deck chart. One-deck S17 A,7 vs ace is a stand on our table and is not marked.
• Split 9s vs 9, but stand vs 7. Source: Wizard 4–8 deck chart (“Split 9s against a dealer 2–6 or 8–9”). Our tables agree.
• Don’t surrender 8s vs 10 — split them. Surrendering hard 16 vs 10 is right; a pair of 8s is not treated as that 16. Source: 4–8 deck chart and the Wizard surrender page. One published exception: with late surrender and a dealer who hits soft 17, 8s vs ace is surrender-else-split.
• 11 vs ace in a six- or eight-deck S17 game is a hit, not a double. Source: 4–8 deck chart. H17 shoes, and 1- or 2-deck S17, still double — those are not marked with this line.
• Don’t split 10s vs 6. 20 is already a strong total. Source: 4–8 deck chart (“Never split 5s and 10s”).

The calculator’s Random button can prefer these sparkle hands when that option is on.

A minimum margin (proposal)

Some exceptions win by a sliver. As a proposal for Jake (not yet a locked decision), the calculator hides an exact-card note — and follows the total-dependent chart — when the EV margin is under 0.0005 units (five hundredths of a cent per dollar bet). Under the default six-deck S17 DAS, no-surrender set, 7 of 101 exceptions fall below that cutoff, including 7,8,ace vs 10 (margin +0.000355) and 2,4,10 vs 10 (+0.000134). The well-known 10,2 vs 4 stays visible (+0.000751). Flagged for Jake: a different cutoff, or “always show the exact-card play,” is easy to apply.

10,2 versus a dealer 4

The six-deck S17 DAS chart says stand on hard 12 vs 4. That is the frequency-weighted action, and it matches both Wizard of Odds and Beating Bonuses (350/350 on this table, 10 Nov 2026). The four two-card compositions do not all agree:

Hit is better for 10,2 by 0.000751 units. The other three holdings prefer stand, and they occur often enough that the chart stands. On one or two decks the same 10,2 exception is larger (one-deck margin +0.017883; two-deck +0.007394); both were verified against the Wizard of Odds hand calculator on 10 Nov 2026.

Three-card 16 versus a 10

The two-card chart says hit on hard 16 vs 10 when surrender is not offered — a close call (weighted hit −0.534819, stand −0.540125). After a third card, many six-deck holdings prefer to stand, because the small cards you would have wanted to draw are already in your hand. Examples under default rules, all verified on 10 Nov 2026 against the Wizard of Odds hand calculator, with EVs agreeing to the reported six decimals:

A pair of eights vs 10 is not in this list. It is a split, not a 16. See 16 vs 10 and splitting 8s.

Multi-card soft 18 vs ace

Two-card A,7 vs ace is a hit under six-deck S17 (hit −0.095345, stand −0.100278). Some later holdings — for example 3,3,A,A and 2,2,3,A — prefer to stand. Both of those were verified on 10 Nov 2026 (stand EVs −0.102773 and −0.103558 respectively).

How far we enumerate

Two-card hands: every composition, full action set. Three to six player cards: hard 12–17 and soft 17–19 only, hit versus stand. After six cards, or outside that window, the calculator falls back to the chart. Notes marked verified were spot-checked on the Wizard of Odds hand calculator. Notes marked computed have not been. Across 32 rule-chip combinations the engine found 5,436 exceptions (96 two-card, 5,340 multi-card). Under the default rule set there are 101.

Method summary: finite-shoe expected values, US peek, the independence approximation for non-ace splits described on About accuracy. These notes are still basic strategy, not counting.