Baccarat is the casino game most often described as 'almost a coin flip.' Two hands — Banker and Player — get cards, and whoever gets closer to 9 wins. There's no skill, no strategy, no decisions for the gambler to make. Just bet, watch, win or lose. Even-money payouts on both sides.
Almost a coin flip. Not quite. Banker wins about one more bet per hundred than Player. The asymmetry is small enough that players don't see it in a single session but big enough that it's the entire baccarat business model. This article explains where that one-in-a-hundred difference comes from.
The game in 60 seconds
Player and Banker each get two cards. Card values: 2-9 face value, 10/J/Q/K count as 0, Ace counts as 1. The hand's score is the sum modulo 10 — so a 7 plus 8 is 15, which becomes 5. Closest to 9 wins. A natural is an initial two-card 8 or 9 — game ends, no third card.
If neither hand is a natural, the third-card rules decide whether each side draws. The rules are fixed — no decisions to make — and asymmetric between Player and Banker. That asymmetry is the entire game.
The third-card rule (and why Banker draws based on Player's card)
Player goes first. Player draws a third card whenever the initial total is 0-5; stands on 6-7. Simple, no information advantage.
Banker is more complex. Banker's third-card rule depends on Player's third card (if Player drew one):
- Banker total 0-2: always draw
- Banker total 3: draw unless Player's third was 8
- Banker total 4: draw if Player's third was 2-7
- Banker total 5: draw if Player's third was 4-7
- Banker total 6: draw if Player's third was 6-7
- Banker total 7: stand
Banker is the second mover with information about what Player drew. That information advantage — small but real — is why Banker wins more often than Player. The casino isn't cheating; the rules are literally giving Banker a peek.
The win rates: 45.9% / 44.6% / 9.5%
Across an 8-deck shoe (standard) with full simulation:
- Banker wins: 45.86% of hands
- Player wins: 44.62% of hands
- Tie: 9.52% of hands
Banker beats Player by 1.24 percentage points — not because Banker is luckier but because Banker's draw rules let it adjust to Player's third card. A bet on Player resolves with no commission. A bet on Banker resolves with a 5% commission on winnings, paid to the house. That commission is how the casino keeps Banker from being a positive-EV bet.
The 5% commission and what it costs
Without commission, Banker would return more than 100% in expectation — it'd be a profitable bet for the player. The casino can't let that stand, so they tax Banker wins at 5%. Math:
- Banker wins 45.86% of hands at 0.95:1 (after 5% commission)
- Banker loses 44.62% of hands at -1:1
- Tie 9.52% — push, no money changes hands on Banker bets
- EV per $1 Banker bet: 0.4586 × $0.95 − 0.4462 × $1 = $0.4357 − $0.4462 = −$0.0106
- House edge: 1.06%
Player bet (no commission):
- Player wins 44.62% at 1:1
- Player loses 45.86% at -1:1
- Tie 9.52% — push
- EV: 0.4462 × $1 − 0.4586 × $1 = −$0.0124
- House edge: 1.24%
Banker is 0.18 percentage points cheaper than Player. Always bet Banker. The standard baccarat strategy fits on a Post-it.
Why you should never bet Tie
Tie pays 8:1 (sometimes 9:1, rarely). The actual probability of a tie is 9.52%, so the fair payout would be about 9.5:1.
- Tie at 8:1: probability 0.0952, payout 8, expected return = 0.0952 × 8 − 0.9048 = -0.144
- House edge on Tie: 14.4%
- Tie at 9:1: house edge drops to about 4.85% — still terrible
The streak illusion: baccarat is memoryless
Every baccarat table comes with a scoreboard showing the recent results — usually a grid of red and blue dots indicating Banker and Player wins. Players study these grids religiously, looking for streaks, breaks, and 'patterns' that predict the next hand.
The deck has memory (cards leave the shoe) but the influence is tiny. The board is, for practical purposes, a coin-flip log. A 'streak' of 6 Banker wins doesn't make the 7th more likely (gambler's fallacy) or less likely (regression-to-mean fallacy). The probability of each hand is still 45.86% Banker, 44.62% Player, regardless of what came before.
Players who bet 'with the streak' or 'against the streak' get exactly the same expected loss as players who pick at random. The board is theater; the math is unchanged.
Card counting baccarat: real, tiny, mostly pointless
Yes, you can count cards in baccarat. The math has been worked out exhaustively — the most famous study is by Peter Griffin in 'The Theory of Blackjack' (also true for baccarat). The conclusion: with a perfect count and deep penetration, the player can earn approximately $0.70 per hour at $25 minimums. Not $70, not $700. Seventy cents.
The reason is that the third-card rules are so close to optimal that the deck composition barely matters. A few specific cards (sevens, eights, nines) shift the edge by tenths of a basis point. By the time you reach a meaningful count advantage, the dealer is shuffling the shoe.
Some advantage players have used baccarat as a 'comp generator' — flat-betting Banker for a tiny edge while collecting cash-back and free hotel rooms. The actual edge is in the comps, not the game. If that math interests you, see the Comp Rate post on this blog.