GuideStaking plans, simulated
Baccarat Betting Systems: Do They Work? Eight Plans, a Million Shoes
Do baccarat betting systems work, or is every baccarat strategy built on staking just a way to lose at a different speed? Baccarat asks one choice of you before the cards are drawn, and the only choice that affects your bankroll is where your chips go. To answer the question with numbers, we ran eight staking plans through the same simulated shoes.
What a betting system can and cannot change
The cost of playing baccarat is fixed in the rules and the payouts. A betting system decides how much you stake on the next hand; it never decides what that hand costs.
When you back Banker at a standard table, the house edge is 1.06%: the casino expects to keep £1.06 for every £100 you stake over the long term. Player costs 1.24%. No betting plan changes either percentage. The staking plans you will find in books and forums tell you whether to bet £10 or £50 on the next hand, based on what happened in the last ones. Changing your stake alters how your results swing and how deep a losing run can go, but it never moves the edge by a fraction.
Over enough hands, Banker results drift around what the edge predicts, nothing more. A baccarat betting system only manages your money while the edge works on every pound you put down. Staking more after a loss, or less after a win, changes when and how you hand over the expected loss, not its size per pound staked.
The same holds online and in a card room: the staking plans are identical, and so is the edge underneath them. The results below show it. Every plan was played on the same shoes, and every plan that bets Banker comes out close to the Banker edge per £100 staked. What differs is how often a session ends ahead, and how bad the worst ones are.
That is why the question to ask of any baccarat strategy built on staking is not whether it wins, but what it costs and how it fails. The odds and strategy page sets out the cost of each bet; this page measures the failures.
Eight staking plans, a million shoes each
Over 1,000,000 shoes, every plan on Banker returned close to −£1.06 per £100 staked, and following the last winner landed between Banker and Player.
Flat £10 on Banker returned −£1.06 per £100 staked, against an exact edge of 1.06%. The Martingale returned −£1.07, the Fibonacci −£1.03, the Paroli −£1.05. The small differences between them are the noise of the simulation, not a skill in the plan.
The share of sessions that ended in profit is where the plans differ. The Martingale ended 91.9% of shoes ahead, against 47.2% for flat Banker. It pays for that with the losses in the next table.
| Plan | Total staked | Return per £100 staked | Sessions in profit |
|---|---|---|---|
| Flat £10 on Banker | £812,799,680 | −£1.06 | 47.2% |
| Flat £10 on Player | £812,799,680 | −£1.23 | 43.1% |
| Martingale (Banker) | £3,376,589,140 | −£1.07 | 91.9% |
| Fibonacci (Banker) | £2,403,708,110 | −£1.03 | 82.5% |
| D'Alembert (Banker) | £3,918,659,530 | −£1.06 | 64.4% |
| Paroli (Banker) | £1,393,255,760 | −£1.05 | 44.8% |
| 1-3-2-6 (Banker) | £1,625,518,730 | −£1.07 | 44.9% |
| Follow the last winner | £812,799,680 | −£1.15 | 45.5% |
Own simulation: 1,000,000 eight-deck shoes per system (the same shoes for each), seed 20261106, run 6 November 2026. Base stake £10, table limit £5,000, one session = one shoe of about 81 hands.
Martingale baccarat: the doubling plan and its losing run
The Martingale doubles your stake after every loss and goes back to the base stake after a win. It is the most famous plan at the table and the one with the worst failures.
Start with £10 on Banker and lose, and the next bet is £20. A losing run of seven hands costs £10, £20, £40, £80, £160, £320 and £640: £1,270 in all, and the eighth bet is £1,280 to come out £10 ahead. On Banker the 5% commission means a win on £1,280 pays £1,216, so even the recovering win leaves the run £54 down. Running the doubling on Player avoids the commission but carries the higher edge of 1.24%.
Every table has a maximum stake. With a £5,000 limit, the ninth decided loss in a row leaves you £5,110 down and unable to place the £5,120 the plan asks for. In the simulation that happened 58,310 times in 1,000,000 shoes, about once every 17 shoes. The median shoe ended £284 up; the worst ended £15,236.50 down.
You cannot outrun a bad streak by throwing doubling amounts of money at a 1% disadvantage. The bankroll the Martingale needs grows with every loss, and the crash, when it comes, takes back hours of small, steady gains in one sequence.
How the sessions ended: the typical shoe and the worst
A plan that wins most shoes has to lose the rest heavily, because the total cost per £100 staked does not move.
The median shoe tells you what a typical evening looks like; the next column is the line that one shoe in twenty finished below; the worst session and the deepest drawdown show where each plan can take you. Flat Banker's worst shoe lost £388.50; the Martingale's lost £15,236.50, and its deepest fall from a high point within one shoe was £16,476.
| Plan | Median shoe | 1 shoe in 20 ends below | Worst shoe | Deepest drawdown |
|---|---|---|---|---|
| Flat £10 on Banker | −£8.50 | −£145.50 | −£388.50 | £388.50 |
| Flat £10 on Player | −£10 | −£150 | −£440 | £450 |
| Martingale (Banker) | £284 | −£4,781.50 | −£15,236.50 | £16,476 |
| Fibonacci (Banker) | £92 | −£201.50 | −£20,159 | £20,178.50 |
| D'Alembert (Banker) | £151.50 | −£1,100.50 | −£7,342.50 | £7,349.50 |
| Paroli (Banker) | −£21.50 | −£281 | −£572.50 | £582.50 |
| 1-3-2-6 (Banker) | −£27 | −£332.50 | −£704.50 | £720 |
| Follow the last winner | −£9.50 | −£148 | −£425.50 | £425.50 |
Own simulation: 1,000,000 eight-deck shoes per system (the same shoes for each), seed 20261106, run 6 November 2026. Base stake £10, table limit £5,000, one session = one shoe of about 81 hands.
Fibonacci baccarat, D'Alembert and other progressions
Because the Martingale is so aggressive, many players look for a progression that climbs the staking ladder more slowly.
The Fibonacci system takes its stakes from the sequence 1, 1, 2, 3, 5, 8, 13, 21 and so on, each number the sum of the two before it. You move up one step after a loss and back two steps after a win. The stake still climbs with each loss until the table maximum stops it. It feels less aggressive than doubling, but it still needs a much larger bankroll than flat betting to survive a poor run of cards.
In the simulation the Fibonacci on Banker returned −£1.03 per £100 staked and hit the £5,000 limit 8,267 times. Its worst shoe ended £20,159 down, worse than the Martingale's, because the Fibonacci keeps climbing for longer before the limit stops it.
The D'Alembert is flatter still: one unit more after a loss, one unit less after a win. It rests on the idea that wins and losses balance out in the short term. The shoe has no memory of what happened in the last hand. The plan returned −£1.06 per £100 staked, with 64.4% of shoes ending ahead.
You still pay the 5% commission on every winning Banker bet, which eats into the arithmetic of every recovery system. Progressions organise your losses into a neat row and give you something to concentrate on while the cards are dealt. They do not change what each hand costs.
Paroli and 1-3-2-6: pressing wins
Rather than chase losses, positive progressions raise the stake only after a win.
The Paroli doubles the bet after a win and resets after three wins in a row or after any loss. From a £10 base, three wins bring £70 at even money on Player (10 + 20 + 40); on Banker the commission makes it £66.50. A loss costs the £10 you put in plus whatever winnings from the run were still riding on the table.
The 1-3-2-6 follows the same idea in units: one, then three, then two, then six, moving on only after a win and going back to one after the sixth unit or after any loss. The aim is to bank a small profit if the run breaks part way. In the simulation the Paroli returned −£1.05 per £100 staked and the 1-3-2-6 −£1.07.
These plans feel less stressful than the Martingale, and their worst shoes are far milder. But the casino keeps the same £1.06 of every £100 wagered on Banker, whether the wager was a flat bet or the last step of a successful run. Those who sell a staking plan built on pressing wins show the big payouts and leave out the many times the run breaks on the second or third step.
Baccarat patterns: following streaks and trends
The other family of systems ignores stake size and tries to predict the next hand from the last ones.
Players study the bead road and the big eye boy on the display, looking for streaks and repeating baccarat patterns. They wait for a run of Player wins to break, or they bet on the hands to chop back and forth. The cards come from an eight-deck shoe under a fixed drawing rule, and the last hand gives no useful prediction of the next one. The roads record the past; they do not forecast.
The simplest trend plan is to follow the last winner: bet on whichever side won the previous decided hand. Over 1,000,000 shoes it returned −£1.15 per £100 staked, between flat Banker (−£1.06) and flat Player (−£1.23), which is what you would expect from a plan that ends up on each side part of the time.
Baccarat is the cheapest table in the casino if you flat bet Banker and leave it there. Any attempt to predict the shoe or vary your stakes adds complexity to a game decided by the cards. If you want to read what the remaining cards really do to the odds, the card counting page runs the exact numbers, and the rules page shows the chart that decides every draw.
How the shoes were simulated
Every figure on this page comes from our own shoe simulator, run on 6 November 2026 with seed 20261106.
Each of the 1,000,000 shoes is eight decks, shuffled by a seeded random generator so the run can be repeated. The first card is turned and that many cards are burned; the cut card sits 14 cards from the end, and when it comes out the hand in play is finished and one more is dealt. Hands follow the standard drawing chart, Banker wins pay less 5% commission, ties return the stake. All eight plans bet on the same shoes, from a £10 base with a £5,000 table limit, and a session is one shoe. To check the simulator, flat Banker had to come out near the exact 1.06% edge and the win shares near 45.86, 44.62 and 9.52%; both held within simulation error. Limits: unlimited bankroll, no side bets, one table rule set.
Questions about baccarat betting systems
Does the Martingale work in baccarat?
No. It turns many small wins into rare large losses and leaves the edge where it was. In our simulation it hit a £5,000 table limit about once every 17 shoes.
Can you predict baccarat patterns?
No. The cards come from an eight-deck shoe under a fixed drawing rule, and the roads only record past hands. Following the last winner did no better than flat betting.
What is the best baccarat betting system?
Flat stakes on Banker. No staking plan lowers the 1.06% edge, so keeping your bets steady costs the least and keeps losses predictable.
Does the Paroli system work?
No. It caps the loss per run at the first stake, but the edge per £100 staked stays the same.
Is the Fibonacci safer than the Martingale?
It climbs more slowly, so its worst shoes are milder, but the edge is the same and a long losing run still meets the table maximum.
Do betting systems work in live baccarat?
No. A live table deals by the same rules and carries the same edge as an RNG table, so a staking plan does the same there.