Do Baccarat Patterns Work? What 40,000 Simulated Shoes Show
Walk past any baccarat table and you’ll see players studying the scoreboard — the “Big Road” — looking for streaks, chops and repeating shapes before they bet. I wanted to know if those patterns mean anything, so I simulated tens of thousands of shoes and checked every pattern against the cards. The short answer: the patterns are real shapes, but they don’t predict the next hand.
How the simulation worked
- Real dealing, not coin flips: 8-deck shoes with the standard third-card rules and the cut card 14 cards from the end — so card removal, the one thing that makes baccarat different from a coin, is included.
- The scoreboard read like a player reads it: for 2,000 shoes (160,893 hands), each Big Road was drawn as an image and read back with computer vision (OpenCV), and the patterns were detected as shapes on the board.
- A control group: every result was compared with the same number of shoes made from completely independent hands. If patterns carried information, real shoes would differ from the control.
- A big betting test: 40,000 more shoes, betting each pattern on the hand right after it appears, at casino payouts.
Patterns appear exactly as often as chance predicts
| Pattern | Per shoe | Shoes with at least one | Real vs independent |
|---|---|---|---|
| Streak (4+ in a row) | 4.5 | 99.5% | 1.02× |
| Dragon (7+ in a row) | 0.56 | 44.6% | 1.05× |
| Chop-chop (4 single wins alternating) | 1.33 | 75.7% | 1.03× |
| 1-2 repeated | 0.53 | 42.1% | 1.02× |
| 3-2-1 | 0.58 | 44.4% | 1.00× |
Almost every shoe has a streak and nearly half have a dragon — which is why they feel meaningful. But they show up just as often in shoes of completely independent hands. Half of all runs on the board (50.7%) are a single win, a quarter (24.8%) are two, and 12.2% are three — the same as the control.
Does a streak keep going?
If streaks had momentum, a run that has reached 4 or 7 should be more likely to continue. It isn’t:
| Run has reached | Chance it continues |
|---|---|
| 1 hand | 49.3% |
| 3 hands | 50.1% |
| 4 hands | 49.5% |
| 7 hands (a dragon) | 48.4% |
Every value is within the margin of error of a coin-flip-like 49–50%. A statistical test of the whole board (which run length follows which) found no detectable memory at all — real shoes were indistinguishable from independent hands.
What happened when we bet the patterns
40,000 shoes, betting on the hand right after each pattern completes (ties push and are left out):
| Strategy | Bets | Win rate | Result per bet |
|---|---|---|---|
| Ride the streak | 169,958 | 50.05% | −1.11% |
| Ride the dragon | 20,241 | 49.69% | −1.79% |
| Keep chopping | 50,277 | 50.17% | −0.94% |
| 1-2 keeps going | 19,709 | 50.18% | −0.93% |
| 3-2-1 then chop | 22,347 | 49.82% | −1.63% |
| All pattern bets | 370,227 | −1.18% |
Every strategy lost. Counting only decided hands, flat-betting Banker loses about 1.17% per bet and Player about 1.36% — and the patterns’ combined −1.18% falls right between them. That’s exactly what you’d get by picking Banker or Player at random. The differences between strategies are within the statistical noise of the sample sizes.
Why patterns can’t work
Each hand is dealt from what’s left in a shuffled shoe. Removing a few cards shifts the odds of the next hand by tiny fractions of a percent — far too little to overcome the house edge, and not in a way the Banker/Player scoreboard can show you. The board records what happened; it doesn’t carry information about what comes next.
What actually decides your result
Three things you control — and none of them is the pattern:
- Which bet: Banker has a 1.06% house edge, Player 1.24%, and Tie at 8 to 1 a painful 14.36%.
- Bet size and number of hands: your expected loss is bet × hands × house edge. $25 on Banker for 200 hands is $5,000 wagered and an expected loss of about $53.
- How long you play: in a short session luck dominates — in that 200-hand example the typical result ranges from about −$700 to +$590, and you have roughly a 44% chance of being ahead. Over 10,000 hands that chance falls to about 13%. Play less, and the house edge has less to work on.
Try your own numbers in the baccarat odds & house edge calculator.
Gambling should be entertainment with money you can afford to lose. If it stops being fun, the U.S. National Problem Gambling Helpline is 1-800-GAMBLER (call or text). This article is math education; it doesn’t promote any casino.
Frequently asked questions
Do baccarat patterns predict the next hand?
No. In a simulation of 40,000 eight-deck shoes, streaks, chops, dragons and other Big Road patterns appeared exactly as often as in independent hands, and betting on them lost 1.18% per bet — the same as betting Banker or Player at random.
Is a baccarat streak likely to continue?
No more than any other hand. In the simulation a run that had reached 4 continued 49.5% of the time and a 7-hand dragon 48.4% — both within the margin of error of an ordinary hand.
What is the best bet in baccarat?
Banker. Even after the 5% commission, its house edge is about 1.06% in an 8-deck game, versus 1.24% for Player and 14.36% for a Tie that pays 8 to 1.
How much will I lose playing baccarat?
On average, bet size × number of hands × house edge. Betting $25 on Banker for 200 hands is $5,000 wagered with an expected loss of about $53, but short sessions swing widely — roughly −$700 to +$590 in 95% of cases.
How often does a dragon (7 in a row) appear in baccarat?
In the simulation, a run of 7 or more appeared in about 45% of shoes — common enough to feel meaningful, but no more often than in a sequence of independent hands.