Does Card Counting Work in Baccarat?

Every week, new baccarat players ask the same question after watching a dealer pull cards from an eight-deck shoe: does card counting work in baccarat the way it works in blackjack? After all, both games deal from a finite multi-deck shoe without replacement until the plastic cut card appears, and both games leave discarded hands in the acrylic discard tray on the left side of the table.

The short mathematical answer—established first by mathematician Edward O. Thorp and computer scientist William E. Walden in the 1960s and confirmed by exact 4.998-quadrillion-deal combinatorial evaluations—is that card counting is theoretically real in baccarat, but practically useless. In an eight-deck (416-card) shoe, removing individual ranks shifts the baccarat house edge by at most 0.0116 percentage points per card—roughly one-fiftieth the impact of removing an Ace or Ten in blackjack. As a result, a linear card counter waiting for the Banker or Player main wager to cross into a positive mathematical edge will find a qualifying hand on fewer than 1 in 1,000 hands (and even with exact full-rank computer tracking across all main and Tie wagers, only a tiny fraction of a cent per unit per hour-long shoe).

Below, we walk through the exact Effect of Removal (EOR) for every card rank from Ace through Ten, explain the three structural reasons Punto Banco neutralizes blackjack-style counting, inspect classic integer point-count tags, and use an 18-hand shoe (BBPBBPPBBBPPBBPBTB) in our free Baccarat Calculator & Road Decoder to show why counting scoreboard beads or baccarat roads has zero connection to card counting.

1. Why card counting works in blackjack—and why baccarat breaks it

To understand why baccarat and blackjack respond so differently to deck depletion, compare the rules that give a blackjack card counter an edge against the rules of Punto Banco third-card drawing:

1. No asymmetric bust penalty: In blackjack, any total over 21 busts immediately, and the dealer must hit stiff hands (12–16) whenever a 10-value card is drawn. A shoe rich in 10s and Aces busts the forced-drawing dealer far more often than a standing player. In baccarat, no hand can ever bust: modulo-10 arithmetic simply drops the tens digit (so 8 + 7 = 15 → 5), and the 128 zero-value cards (10, J, Q, K) add 0 to whichever side draws them.
2. No 3:2 bonus payout on Naturals: In blackjack, a two-card 21 pays 3:2 (1.5×) to the player while a dealer blackjack only collects 1×. That asymmetric payout is the #1 reason Aces and Tens carry huge Effect of Removal values in blackjack. In standard baccarat, a two-card Natural 8 or Natural 9 pays the exact same 1:1 on Player (or 0.95:1 after 5% commission on Banker) as a three-card win of 1 over 0.
3. No mid-hand player decisions: A blackjack player can change strategy based on the true count—standing on 16 against a dealer 10, doubling down on 9 or 10, or splitting pairs. In Punto Banco baccarat, once your wager is placed before the first card leaves the shoe, both the Player hand and the Banker hand follow a 100% mandatory, symmetric-feeling drawing tableau with zero human choices.
4. Banker and Player share almost the same cards: Any card that helps produce a high point total (such as an 8 or a 9) is almost equally likely to be dealt to the Player position or the Banker position. The only mathematical difference between the two sides comes from Banker acting second and knowing Player's third card on totals of 3, 4, 5, and 6.

2. Effect of Removal (EOR) for every card rank in an 8-deck shoe

In combinatorial game theory, the Effect of Removal (EOR) measures how many percentage points a wager's expected value changes when one single card of a given rank is removed from a fresh 416-card shoe (leaving 415 unseen cards). Because a fresh 8-deck shoe starts with Banker at −1.0579% and Player at −1.2351%, a rank's EOR would have to accumulate by more than +1.06 percentage points on Banker (or +1.24 percentage points on Player) just to reach a break-even 0.00% expectation.

Look at the exact single-card EOR shifts across all ten baccarat card values, first derived by Edward Thorp and William Walden and refined via full 8-deck combinatorial enumeration:

Why the 4 and the 6 pull in opposite directions

Notice why the two most influential ranks in the entire deck are 4 (+0.0116% toward Banker when removed) and 6 (+0.0113% toward Player when removed). Both numbers sit right on the seam of the third-card tableau:

• The Player hand draws on two-card totals of 0–5 and stands on 6–7.
• The Banker hand stands on 6 unless Player drew a third card of 6 or 7, and Banker draws on 5 when Player's third card is 4, 5, 6, or 7.

When extra 5s, 6s, and 7s remain unseen in the shoe, Banker benefits slightly because Banker's conditional drawing rules are tuned to catch up against Player totals built from those middle-high ranks. Conversely, when extra Aces, 2s, 3s, and 4s remain unseen in the shoe (meaning 5s, 6s, and 7s have already been dealt and discarded), Player's signed expectation rises slightly.

Why linear level-1 baccarat count tags almost never trigger

Notice the fifth column in the table above: the classic balanced integer count tags (−1 on A, 2, 3; −2 on 4; +1 on 5, 7, 8; +2 on 6; and 0 on 9, 10, J, Q, K). In each 52-card deck, the negative tags sum to −1 − 1 − 1 − 2 = −5 per suit (−20 per deck) and the positive tags sum to +1 + 2 + 1 + 1 = +5 per suit (+20 per deck), forming a balanced running count of 0 across 416 cards.

Yet look at the scale required to overcome the house edge:

• To overcome Banker's −1.0579% starting house edge purely by removing 4s from a fresh shoe, you would have to remove 1.0579 ÷ 0.0116 = 91.2 fours—which is physically impossible because an entire eight-deck shoe only contains 32 fours!
• Under the balanced level-1 count above, the running count divided by the exact number of unseen cards (or unseen decks) must reach an extreme true-count threshold deep in the final deck of the shoe before either Player (when the true count is extremely positive) or Banker (when the true count is extremely negative) crosses 0.00% mathematical expectation.

3. What about counting side bets (Dragon 7, Panda 8, Pair, Tie)?

You may have heard casino lore about advantage teams counting cards at baccarat tables. Whenever academic mathematicians (such as Eliot Jacobson and Michael Shackleford) documented real-world vulnerability in baccarat, it was never on the standard Banker or Player main bets—it was on specific high-payout optional side bets whose rules depended heavily on one or two specific card ranks:

• 8:1 Tie bet and zero-value cards: Because 10, J, Q, K all count as 0, a shoe whose remaining unseen cards are unusually packed with zero-value cards pushes both hands toward identical totals (especially 0-0 ties). However, because the 8:1 Tie bet starts at a massive −14.36% house edge, the unseen shoe must reach an extreme surplus of 10-value cards deep in the shoe just to reach break-even.
• EZ Baccarat Dragon 7 (40:1) and Panda 8 (25:1): Dragon 7 pays 40:1 when Banker wins with a three-card total of 7, while Panda 8 pays 25:1 when Player wins with a three-card total of 8. Because a three-card 7 or 8 requires specific combinations of low third cards and 8s/9s, those side bets are more sensitive to rank depletion than the main bets. Modern casinos counter this automatically by placing the cut card at 14 to 26 cards (or one full deck) from the back of the shoe, dealing face-down burn cards at the start of every shoe, banning mid-shoe entry, or using pre-shuffled sealed shoes and continuous shuffling machines.

4. Why “counting the scoreboard” is not card counting

The most common misconception in casino pits is confusing card counting (tracking the 416 physical card ranks A–9 and 10/J/Q/K) with scoreboard counting (counting how many Red or Blue circles appeared on the five baccarat roads).

To see why a scoreboard grid contains zero information about which card ranks remain in the shoe, examine this 18-hand sequence (BBPBBPPBBBPPBBPBTB—11 Banker wins, 6 Player wins, and 1 Tie on Hand 17):

BBPBBPPBBBPPBBPBTB

What the 5 grids actually record on this 18-hand shoe

Look at what happens across the five scoreboard grids in BBPBBPPBBBPPBBPBTB:

1. Bead Plate & Big Road (Red = Banker, Blue = Player): The 18 hands form 9 Big Road columns with depths 2, 1, 2, 2, 3, 2, 2, 1, 2, plus a green diagonal Tie slash on Column 9, Row 1 (Hand 17). Notice that a Banker win recorded as a red circle could have been won with two low cards (such as 4 + 4 = Natural 8, removing two 4s and shifting the shoe slightly toward Banker) or three middle-high cards (such as 6 + 6 + 6 = 18 → 8, removing three 6s and shifting the shoe slightly toward Player). The Big Road circle never records which card ranks left the shoe.
2. Big Eye Boy (Lookback 1 — 14 entries: B B B R R R B B R B R R B B): On Big Eye Boy, Small Road, and Cockroach Pig, red never means Banker and blue never means Player; red means pattern repetition/symmetry and blue means a pattern break. When Column 3 starts at Row 1 (Entry 1), Rule 1 compares Column 2's depth (1) with Column 1's depth (2) → different depth = Blue. When Column 3 continues to Row 2 (Entry 2), Rule 2 looks one column left at Column 2: because Column 2 ended at Row 1, Row 2 is the first missing row transition → Blue.
3. How Rule 2's three cases drive Entries 4–7 and 14: Remember that Rule 2 (continuing down a column to Row r ≥ 2) has three plain-English cases: if both cells exist in the comparison column (Row r and Row r − 1), the mark is Red; if Row r is the first missing row right past the end of the shorter column, the mark is Blue; and if both comparison cells are blank (two or more rows past a shorter column), the mark is Red. In our shoe, Entries 4 and 6 (Column 4 Row 2 and Column 5 Row 2) both find filled cells at Row 2 of their lookback columns → Red, while Entry 7 (Column 5 Row 3) steps one row past Column 4's depth of 2 → Blue, and Entry 14 (Column 9 Row 2) steps one row past Column 8's depth of 1 → Blue.
4. Next-hand preview (Hand 19 — If B vs If P): After Hand 18, Column 9 sits at depth 2. If Hand 19 is Banker (continuing down Column 9 to Row 3 via Rule 2), Big Eye Boy checks Column 8 (depth 1): because both Row 2 and Row 3 are blank in Column 8 (Rule 2's third case: both comparison cells blank), Big Eye Boy records Red, while Small Road (checking Column 7, depth 2, first missing row at Row 3) records Blue and Cockroach Pig (checking Column 6, depth 2, first missing row at Row 3) records Blue. If Hand 19 is Player (starting Column 10 at Row 1 via Rule 1, comparing Column 9's finished depth of 2), Big Eye Boy compares against Column 8 (depth 1) → Blue, while Small Road (Column 7, depth 2) and Cockroach Pig (Column 6, depth 2) both match 2 = 2 → Red · Red.

That entire derived-road sequence is a pure geometric summary of column heights—which is why past hands and road patterns never predict the next hand.

5. Where to focus instead of counting cards

Once you know that card counting in an eight-deck baccarat shoe yields a positive mathematical expectation on fewer than 1 in 1,000 hands, you can stop wasting mental energy tallying 4s and 6s at the table and focus on the two things that actually matter:

• Understand the fixed cost of every bet on the felt: Comparing standard Banker (~1.06% house edge) and Player (~1.24% house edge) against the 8:1 Tie (~14.36% house edge) saves more expected value on a single coup than 100 hours of eight-deck card counting ever could.
• Build effortless scoreboard and third-card fluency: Learn how Baccarat Smart works, try our free interactive lesson preview, or explore the full Baccarat Smart course curriculum so you can read any Macau tote board at a glance for pure enjoyment.

Test this 18-hand shoe in the live Road Decoder below

Use the interactive shoe below to tap + Banker (and watch Big Eye Boy record Red via Rule 2's both-cells-blank case while Small Road and Cockroach Pig turn Blue) or + Player on BBPBBPPBBBPPBBPBTB, or open the full eight-road free Baccarat Calculator & Road Decoder to inspect the six D3 statistical randomness charts live.