Baccarat Card Counting: Can It Actually Beat the Game?
Baccarat card counting sounds like it should work. Cards are dealt from a shoe without being replaced, so every card that leaves changes the mix that remains — exactly the property that makes counting possible in blackjack. The question is not whether the odds move, but by how much, and how often they move far enough to matter. We answered that by computing the exact odds of an 8-deck shoe (416 cards) with each card rank removed, then dealing thousands of simulated shoes. The short answer: on the Banker and Player bets, counting is real but almost worthless; on the Tie it is noisy; on the Pair side bet it can occasionally flip the edge, if you track every rank and the shoe is dealt very deep.
Why counting works in blackjack but barely in baccarat
In blackjack, the player makes decisions — stand, hit, double, split — and a deck rich in tens and aces changes the best decision and pays 3 to 2 for blackjacks more often. Both effects push in the same direction, so a simple +1/−1 count captures most of the value. In baccarat nobody makes a decision after the bet: both hands draw by a fixed table. And the two main bets are mirror images. Any card that helps Banker hurts Player by almost the same amount, so the cards largely cancel out over a shoe. On top of that, the house edges you would need to overcome are already small (Banker 1.06%, Player 1.24%), yet the shift from each card is far smaller still.
Effect of removal: what one card does in an 8-deck shoe
The table below shows how much the expected value of each bet changes when one card of a given rank is taken out of a full 8-deck shoe. Values are in percentage points of the bet; a plus sign means the bet became better for the bettor.
| Card removed | Banker | Player | Tie (8 to 1) |
|---|---|---|---|
| A | +0.0044 | −0.0045 | +0.0129 |
| 2 | +0.0052 | −0.0054 | −0.0239 |
| 3 | +0.0065 | −0.0067 | −0.0214 |
| 4 | +0.0116 | −0.0120 | −0.0292 |
| 5 | −0.0083 | +0.0084 | −0.0264 |
| 6 | −0.0113 | +0.0113 | −0.1160 |
| 7 | −0.0083 | +0.0082 | −0.1091 |
| 8 | −0.0050 | +0.0053 | +0.0654 |
| 9 | −0.0023 | +0.0025 | +0.0426 |
| 10, J, Q, K (each) | +0.0019 | −0.0018 | +0.0513 |
Three patterns stand out. First, small cards (A–4) leaving the shoe favour Banker, while 5–9 leaving favour Player. Second, the 4 is the most powerful card for Banker and the 6 for Player, but even the 4 moves Banker by only about one hundredth of one percent. Third, the Tie bet reacts most strongly to 6s and 7s — logical, because 6-6 and 7-7 are the two most common tie totals (41.6% of all ties together); when those cards are gone, ties become rarer.
How many cards would it take? A worked calculation
Banker starts at −1.0579%. Removing one 4 improves it by about 0.0116 points. To bring Banker to break-even with 4s alone you would need about 1.0579 ÷ 0.0116 ≈ 91 fours to have left the shoe — but an 8-deck shoe contains only 32. Even combining every helpful rank at once, the remaining cards have to be extremely lopsided, late in the deal, before Banker or Player turns positive. The effect of each card grows as the shoe shrinks (one card out of 52 matters eight times more than one out of 416), so whatever chances exist come in the last deck — and by then the shoe is nearly over.
Shoe simulation: how often is a main bet actually positive?
We dealt 3,000 random 8-deck shoes for each scenario, tracking every rank perfectly, and estimated each bet's value before every hand with the removal effects above (a linear approximation; for Banker and Player it stayed within about ±0.25 points of the exact calculation in our checks). The only variable that mattered was penetration — how many cards are left behind the cut card when the shoe ends.
| Cut card position | Hands per shoe | Banker positive | Player positive | Tie positive |
|---|---|---|---|---|
| Last 52 cards not dealt | 74.1 | 0.000% of hands | 0.001% | 0.000% |
| Last 14 cards not dealt | 81.8 | 0.148% (0.12 per shoe) | 0.057% (0.05 per shoe) | 0.013% (0.01 per shoe) |
Read the bottom row carefully: even with deep penetration and perfect tracking, a positive Banker bet appears about once every eight shoes, and when it does the average advantage is about 0.20%. Betting 100 units on that one hand is worth about 0.2 units — after sitting through roughly 650 hands waiting for it. The Tie is occasionally positive by a larger margin (about 1.8% on average) but only about once in a hundred shoes. In practical terms, counting the main bets does not overcome the house edge.
The exception: counting the Pair side bet
The Player Pair and Banker Pair bets are different, because their odds depend directly on how many cards of each rank remain. With a fresh 8-deck shoe the chance of a pair is 7.47%, and an 11 to 1 payout gives a 10.36% house edge. The bet becomes positive when the pair probability exceeds 1 in 12, or 8.33%. That needs the remaining cards to be concentrated in a few ranks, which happens late in a shoe. We calculated the pair probability exactly before every hand of 20,000 simulated shoes:
| Cut card position | Pair bet positive | Per shoe | Average edge on those hands | Highest pair chance seen |
|---|---|---|---|---|
| Last 52 cards not dealt | 0.531% of hands | 0.39 hands | +3.75% | 11.46% |
| Last 14 cards not dealt | 2.226% of hands | 1.82 hands | +11.06% | 23.81% |
So the Pair bet is the one place where tracking cards produces a measurable edge — about 0.20 units per shoe in the deep-penetration case if you bet one unit on each positive hand, and about 0.015 units per shoe with a one-deck cut. Notice what it requires: a separate count for all 13 ranks (a pair is about rank, so a 10 and a king do not pair), recalculated every hand, at a table that deals almost to the end. That is computer bookkeeping, not a mental count, and the profit per shoe is small compared with the normal swings of an 11 to 1 bet. You can easily see no pair at all on the handful of hands where the bet is positive.
Practical obstacles
- Eight decks dilute everything. Each card is 1/416 of the shoe; most of the shoe is over before the mix drifts noticeably.
- The cut card. Many tables leave a deck or more undealt — the top row of both tables above. That alone removes almost all opportunities.
- Reshuffles. Some games shuffle more often or use continuous shufflers, which reset any count to zero.
- Waiting time. Playing only positive hands means sitting out hundreds of hands; betting every hand to stay at the table costs the normal 1.06–1.24% on all of them.
- House rules. Venues can limit side bets or refuse play to people they believe are tracking cards. Check the rules where you are; this page explains the maths only.
What card counting can and cannot tell you
Counting tells you, precisely, how the remaining cards are composed. It does not predict who wins the next hand in any useful way, and it says nothing about streaks or patterns on the scoreboard — those reflect past results, not the cards left. Even a perfect count changes Banker's win probability by hundredths of a percent. For the underlying numbers see baccarat odds and house edge; for why streak-based systems fail see do betting systems work?. A baccarat predictor shows what the roads would look like after each possible outcome — a good way to see that the pictures change while the probabilities barely do.
Card counting FAQ
Can you count cards in baccarat? Yes. The shoe is dealt without replacement, so the odds shift as cards leave. The shift on Banker and Player is tiny — around 0.01 percentage points per card in an 8-deck shoe.
Is there a simple +1/−1 count for baccarat? You can weight cards by the table above (low cards toward Banker, 5–9 toward Player), but in our simulations even perfect tracking made Banker positive on only 0.148% of hands with deep penetration, and essentially never with a one-deck cut.
Which baccarat bet is most affected by counting? The Pair side bet, because it depends on how many cards of each rank remain. It can turn positive late in a deeply dealt shoe.
Does counting help with the Tie bet? Ties become less likely as 6s and 7s leave and slightly more likely as 8s, 9s and tens leave, but the Tie bet starts 14.36% behind at 8 to 1 and was positive on only about one hand per hundred shoes.
Is card counting illegal? Using your own memory is generally not a crime, but using devices at a table may be, and casinos can refuse play. Laws differ by country.
If you want to see how a shoe actually unfolds, the free Baccarat Scoreboard app records each hand with a tap and marks naturals, pairs and other events — a simple way to see over many shoes how rarely anything unusual really happens.
Effect-of-removal values from exhaustive enumeration of an 8-deck shoe with one card removed; shoe frequencies from simulation (3,000 shoes per main-bet scenario, linear approximation; 20,000 shoes per Pair scenario, exact) — Baccarat Terminal, October 2026. 18+. Informational only; not a gambling service.