Baccarat strategies tested against the maths

Search "baccarat strategy" and you will find systems promising to beat a game that has no decisions in it. This guide does something different: it takes the exact probabilities of punto banco and shows what each popular approach actually does to your expected result. The conclusion is short — bet banker, flat, for a pre-set number of hands — but the reasoning is worth understanding, because it applies to every fixed-odds game you will ever play.

Strategy 1: bet banker, every time

The banker hand wins 45.86% of hands, the player hand 44.62%, and 9.52% tie (eight decks). After the 5% commission, banker costs 1.06% of each stake on average; player costs 1.24%. That 0.18-percentage-point gap is small per hand but real: over 1,000 hands at £25 a hand it is £45 of extra expected loss for choosing player. There is no situation — no streak, no scorecard pattern, no shoe depth — in which player becomes the better bet under standard rules.

£6.36

Expected loss per hour betting £10 banker, 60 hands/hour

£7.44

Same stake and pace on player

£86.16

Same stake and pace on tie at 8:1

The only exception is rule variants. At a "no-commission" table where a banker 6 pays half, banker's edge becomes 1.46% and player (1.24%) is the better bet. At a table charging 4% commission banker drops to 0.60% and becomes even more attractive. Read the rules panel first; you can check any combination in the baccarat chance calculator.

Strategy 2: never bet the tie, and treat side bets as entertainment

The tie pays 8:1 but the true odds against it are about 9.5:1, which gives the house 14.36% — thirteen times the banker edge. At 9:1 it falls to 4.84%, still four times worse than banker. Side bets are the same story with prettier names. Every figure below is for eight decks; the pair bets get worse with fewer decks.

Paytables vary between operators, so the exact edge on a given table may differ by a point or two, but the ranking does not: nothing on the side layout is cheaper than the main banker bet. If you want the thrill of a 25:1 hit, budget for it the way you would budget for a slot session, and keep it separate from your main stake.

Betting systems debunked with the actual numbers

A betting system changes your stake based on previous results. The expected value of each bet is stake × edge, and because every hand is independent, the expected value of a sequence of bets is simply the sum — negative, whatever pattern you follow. What systems really do is reshape the distribution: many small wins and rare catastrophic losses (Martingale), or the reverse (Paroli). Here is what that looks like in baccarat.

Martingale: double after every loss

Start at £10 on banker, double after each loss, return to £10 after a win. Ignoring ties, banker loses a resolved hand 49.32% of the time. The probability of losing seven in a row is 0.4932⁷ ≈ 0.71%, or roughly once every 141 sequences; eight in a row is 0.35%, once every 285. A seven-loss run requires stakes of 10, 20, 40, 80, 160, 320 and 640 — £1,270 lost — and the next bet would be £1,280, which exceeds the £1,000 maximum on most £10 tables. At a typical live table you complete perhaps 30 sequences an hour, so the run that wipes out a week of £10 wins arrives, on average, within five hours of play. Each sequence that ends in a win nets £10 minus the 5% commissions along the way; the average loss per sequence, including the busts, is exactly the house edge times the average total staked. Nothing has changed except the variance.

Fibonacci, d'Alembert and Labouchère

These are slower Martingales. Fibonacci raises the stake along the sequence 1, 1, 2, 3, 5, 8, 13, 21… after losses and steps back two places after a win; d'Alembert adds one unit after a loss and subtracts one after a win; Labouchère crosses numbers off a list. All three need a winning streak to recover a losing one, and in baccarat a win is less likely than a loss on every single hand. The progression is gentler, so the ruin comes later and the bankroll requirement is smaller, but the expected loss per pound wagered is still 1.06% on banker. A d'Alembert player who is 20 units "behind" needs 20 more wins than losses to get level — an event that becomes less likely, not more, the longer they play.

Paroli and other positive progressions

Paroli doubles after a win (usually for three wins, then resets). It cannot blow up the way Martingale can because you are only ever risking the base stake plus house money. Three banker wins in a row happen with probability 0.5068³ ≈ 13% of sequences (ignoring ties); when it hits you win 7 units before commission, and the other 87% of sequences lose 1, 1 or 2 units. Expected value: still −1.06% of total stakes. It is a reasonable way to structure a session if you enjoy streaks — just know it is a structure, not an edge.

Pattern betting on the roads

"Follow the shoe", "bet against the chop", "the dragon tail" — every system based on the scoreboard assumes that past results change future probabilities. They do not. The banker probability after eight consecutive banker wins is 45.86% plus or minus a rounding error caused by the handful of cards removed from the shoe. A simple proof: if streaks were predictable, the casino would not print them on a four-foot screen beside every table. Casinos display the roads because they keep players seated and betting.

Strategy 3: flat betting and 1-3-2-4 as variance management

Once you accept that the expected loss per hand is fixed, the honest question becomes: what shape of results do I want? Two approaches are defensible.

Flat betting

Bet the same amount every hand. Your result after N hands is close to −1.06% × N × stake, with a standard deviation of roughly 0.93 × stake × √N. For 100 hands at £10 that is an expected loss of about £10.60 with a one-standard-deviation range of about ±£93. Flat betting gives you the lowest variance per pound wagered, the most predictable session cost and the simplest bookkeeping. It is what professional advantage players do at any game where they do have an edge, which tells you something.

1-3-2-4

Stake 1 unit; if it wins, stake 3; if that wins, stake 2; if that wins, stake 4; then reset. Any loss resets to 1. The appeal is that after the first two wins you have banked a profit (1 + 3 = 4 units), and the sequence risks mostly winnings. The four outcomes from a fresh start: lose the first hand (−1), win then lose (+1 − 3 = −2), win-win-lose (+4 − 2 = +2), win-win-win-lose (+6 − 4 = +2), win all four (+10). The system does not alter expected value — total stake × 1.06% is what you pay — but it caps the damage on a bad run at 2 units while leaving upside for a streak. Compared with Martingale it is a sensible trade: smaller swings, no table-limit bust, and a built-in stop point after four hands.

Strategy 4: session discipline

Because the hourly cost of baccarat is stake × hands per hour × edge, the two things you control are the stake and the number of hands. A £20 banker bettor playing 70 hands an hour at a live table spends an expected £14.84 an hour; the same bettor on a speed table at 130 hands an hour spends £27.56. Slower is cheaper, and a squeeze table is the slowest of all.

1. 1

   Fix the session bankroll

   Bring an amount you can lose entirely without consequence, typically 1–5% of your total gambling bankroll as set out in our bankroll management guide.

2. 2

   Size the unit at 1–2% of the session bankroll

   A £500 session bankroll means £5–£10 units. This keeps a normal losing streak (8–10 units) from ending the session early.

3. 3

   Set a hand count, not a time

   Decide on, say, 60 hands. Count them on the bead plate. Stop when you reach the number, regardless of result.

4. 4

   Set a stop-loss and a stop-win

   A stop-loss of 50% of the session bankroll and a stop-win of 50% are common. The stop-win is not about locking in 'your' money; it is about leaving before the variance reverts.

5. 5

   Skip the tie, skip the side bets, bet banker

   If the table is no-commission with banker 6 paying half, bet player instead.

6. 6

   Record the session

   Stake, hands, result. After 20 sessions compare your actual loss with stake × hands × 1.06%. It will be close, and that is the whole lesson.

Does card counting work in baccarat?

Only in theory. Edward Thorp analysed it in the 1960s and Peter Griffin later quantified it: with perfect tracking of an eight-deck shoe, player-favourable bets on banker or player arise so rarely, and with so small an advantage, that the expected gain is a few pence per hour at stakes where the casino would already be watching you. The tie and some side bets are more vulnerable late in the shoe, and a small number of professional teams have exploited dealer errors and edge-sorted cards, but that is not strategy available to a recreational player. Blackjack remains the game where counting pays; see our blackjack strategy guide for how it works there.

Summary: the only baccarat strategy that survives the maths

• Bet banker at standard tables; bet player at no-commission tables where banker 6 pays half.
• Never bet the tie at 8:1; treat side bets as entertainment with a separate budget.
• Flat bet, or use 1-3-2-4 if you want streak upside with limited downside.
• Ignore the roads; they record, they do not predict.
• Play fewer hands per hour, at a unit size of 1–2% of your session bankroll, to a fixed hand count.
• Expect to lose about 1% of everything you stake on banker, and treat that as the price of the entertainment.