- Maths reviewed by Josh Miller
- Tested by Diana Westfield
Baccarat card counting works in theory and barely pays in practice. The CasinoCalcs baccarat card counting calculator above shows why, using the same maths a casino’s game protection team would. It deals hundreds of 8-deck shoes, tracks the cards, and tells you how often the count says “bet” and what you’d make. Leave it on Dragon 7 to start: that’s the one bet where a free, simple count does find an edge.
Baccarat card counting on the main bets
Switch the calculator to Banker and set the method to “Perfect count”. That’s the best any system could ever do, using every card seen. Even then, Banker turns positive on a fraction of 1% of hands, and the edge when it does is a couple of tenths of a percent. You’d sit through hundreds of shoes to earn the price of lunch.
The reason is in the table below. Removing one card from a full shoe moves the Banker edge by about a hundredth of a percent. In an 8-deck blackjack shoe a 5 leaving moves it around seven times as much, and blackjack players can change how they play the hand as well. Baccarat’s drawing rules are fixed and both hands get the same cards, so a skewed shoe helps both sides almost equally.
How each card changes the edge
| Card leaving the shoe | Banker | Tie (8:1) | Dragon 7 | Panda 8 |
|---|---|---|---|---|
| A | +0.004 | +0.013 | +0.050 | +0.132 |
| 2 | +0.005 | -0.024 | -0.090 | +0.144 |
| 3 | +0.006 | -0.021 | -0.108 | -0.286 |
| 4 | +0.012 | -0.029 | -0.268 | -0.254 |
| 5 | -0.008 | -0.026 | -0.263 | -0.265 |
| 6 | -0.011 | -0.116 | -0.324 | -0.089 |
| 7 | -0.008 | -0.109 | -0.358 | -0.091 |
| 8 | -0.005 | +0.065 | +0.538 | -0.224 |
| 9 | -0.002 | +0.043 | +0.479 | +0.450 |
| 10, J, Q, K | +0.002 | +0.051 | +0.086 | +0.121 |
Each figure is the change in the bet’s edge, in percentage points, when one card of that value leaves a full 8-deck shoe. Positive means good for you. It’s the “effect of removal”, and every counting system is a rounded version of it.
Where counting does work: Dragon 7 and Panda 8
Look at the Dragon 7 column. Eights and nines are worth +0.5 each when they go, and 4s to 7s are worth −0.3. That’s a big effect, because a three-card Banker 7 needs specific small cards to still be in the shoe. A simple count follows: 4, 5, 6 and 7 count −1, 8 and 9 count +2, everything else 0. Divide by decks left for the true count and bet when it reaches +4.
Run that in the calculator and you’ll see something like this: you bet on about 9% of hands, the average edge on those bets is around +8%, and you’re expected to make about 0.75% of your bet per hand dealt. At £25 and 70 hands an hour, that’s roughly £13 an hour. Before the variance of a 40:1 payout, which is brutal.
Panda 8 is weaker. Our count (tens, pictures, aces and 2s +1, 3s to 5s −2, 6s and 7s −1, 8s −2, 9s +4) needs a true count of +8 and gives a smaller return.
The catch: the cut card
All of this depends on how deep the shoe goes. Set “cards behind the cut card” to 52 (one deck cut off) and the Dragon 7 return halves. Two decks and it nearly vanishes. Positive counts mostly show up near the end of the shoe, and casinos know it.
Also, nobody bets Dragon 7 once every eleven hands without someone noticing. Counting isn’t illegal, but a casino can refuse your side bets.
For the side bets’ starting edges, see the baccarat side bets calculator. If you want counting that really works, the blackjack card counting calculator is a better use of your time.
A count doesn’t make a session safe. The swings on a 40:1 bet are huge. BeGambleAware.org can help if gambling is getting out of hand.