Two opposite claims circulate around baccarat tables, and neither one is quite exact on its own:
- “Every hand in baccarat is completely independent, like flipping a coin.”
- “Because cards come out of a finite eight-deck shoe, scoreboard streaks and card counting give you an actionable edge.”
We ran the exact combinatorial arithmetic across every possible ordered six-card deal in an eight-deck (416-card) shoe—all 4,998,398,275,503,360 deals—calibrated against 1,000 published Wizard of Odds shoes (80,879 hands) and 20,000 simulated casino shoes. Here is what the unseen cards actually change, how tiny that shift really is, and why no scoreboard pattern changes it at all.
1. Exact probabilities of a fresh 416-card shoe
A standard casino shoe holds eight 52-card decks: 32 cards of each rank from Ace through 9, plus 128 zero-value cards (10, J, Q, K). Evaluating every ordered six-card sequence through the official Punto Banco third-card tableau yields the exact baseline:
- Banker wins:
2,292,252,566,437,888of4,998,398,275,503,360deals — 45.86% of all hands (50.68% of decided hands), signed edge −1.06% at 5% commission. - Player wins: 44.62% of all hands (49.32% of decided hands), signed edge −1.24% at 1:1.
- Tie: 9.52% of all hands, signed edge −14.36% at 8:1.
- Hand texture: At least one third card is drawn on 62.1% of hands, and Naturals (8 or 9 on the initial two cards) end 34.3% of hands.
2. How unseen low cards vs high cards nudge the math
Because Player stands on 6–7 while Banker's drawing rules on 3, 4, 5, and 6 depend on Player's third card, removing individual ranks nudges the asymmetry by hundredths of a percentage point:
- Removing a single 4 from a fresh 416-card shoe shifts Banker's signed edge by +0.0116 percentage points.
- Removing a single 5, 6, or 7 shifts Banker's signed edge by −0.0083 to −0.0113 percentage points.
Put simply: when the unseen cards remaining in the shoe are unusually rich in low cards (A–4), the next hand tilts slightly toward Player and third-card draws become more common. When the unseen cards are unusually rich in high cards (5–9), the next hand tilts slightly toward Banker. Zero-value cards (10/J/Q/K) mainly shift Tie frequency.
| Unseen shoe state | Banker (edge) | Player (edge) | Tie (8:1 edge) | ≥1 3rd card | Naturals |
|---|---|---|---|---|---|
| Fresh 416 cards (baseline) | 45.86% (−1.06%) | 44.62% (−1.24%) | 9.52% (−14.36%) | 62.1% | 34.3% |
| 52 unseen, extreme low-rich (A–4 heavy) | 44.90% (−2.56%) | 45.22% (+0.32%) | 9.88% (−11.08%) | 73.3% | 24.3% |
| 52 unseen, extreme high-rich (5–9 heavy) | 46.53% (+0.25%) | 43.94% (−2.58%) | 9.53% (−14.23%) | 58.4% | 36.4% |
| 104 unseen, low surplus from 0-value cards | 45.02% (−1.74%) | 44.51% (−0.51%) | 10.48% (−5.72%) | 61.3% | 34.6% |
Look closely at how extreme and rare those second and third rows are: reaching the final 52 unseen cards with 26 or more low cards remaining occurs in roughly 0.15% of shoes—about 3 shoes in 2,000. Even in that extreme tail, the signed edge barely crosses zero by a quarter to a third of one percentage point.
3. Why card counting in baccarat is far too small to act on
To measure what perfect rank tracking could ever yield in real play, we evaluated exact unseen-card odds before every single hand across 1,000 simulated eight-deck casino shoes (80,887 hands):
- A main or tie bet crossed above
0.00%signed edge on only 0.41% of all hands—roughly 1 hand in 250 (150 Banker instances, 61 Player instances, and 119 Tie instances across 80,887 coups). - Even if a player tracked every exposed rank flawlessly, sat idle for 249 out of every 250 hands, and staked 1 unit solely when an edge turned positive, the total theoretical gain averaged 0.0077 units per entire 81-hand shoe—less than one cent per dollar unit per hour-long shoe.
We also calibrated our composition engine and casino dealer against the published Wizard of Odds 1,000-shoe dataset (80,879 hands): every metric sat well inside normal random variation (|z| < 3: Banker z = −1.45, Tie z = +1.21, Natural z = +0.83, Third card z = −1.20).
4. What about Dragon Tails and scoreboard streaks?
Now compare those tiny card-depletion nudges against what players actually watch on the screen: streaks and derived-road runs. Consider this 24-hand shoe (BPBPBBBBBBBPPBPPBPBBPPBP), where four single-hand chops give way to a seven-hand Banker streak that bends sideways along Row 6 into a Dragon Tail:
Bead Plate
Big Road
Big Eye Boy
Small Road
Cockroach Pig
Does a streak of six or seven wins mean the shoe is “locked into a dragon”—or conversely “due to chop”? Across 20,000 simulated eight-deck shoes, we isolated every single time a shoe produced six or more decided hands in a row on the same side (41,747 observed cases):
- The very next decided hand continued the streak 49.7% of the time (±0.5%).
- A run of three Banker wins reached six wins 13.0% ± 0.07% of the time in a fresh shoe (versus 12.4% ± 0.06% in the extreme 52-card low-rich shoe and 13.6% ± 0.07% in the extreme high-rich shoe). A run of three Player wins reached six 12.0% ± 0.06% fresh (12.6% ± 0.07% low-rich, 11.5% ± 0.06% high-rich).
That is the entire truth in one sentence: earlier hands change the next one only through the cards they used up, and by far too little to act on; no road pattern changes it.
Want to see how a real shoe behaves against fair eight-deck confidence bands? Deal extra hands onto the 24-hand board below or open it in our free Road Decoder to inspect the Fair Shoe Check and Six-Road Randomness Check live.