Baccarat Pattern Strategy — What the Popular Patterns Bet On, and What 8.5 Million Real Hands Say About Them
A baccarat pattern strategy is any rule that reads the scoreboard — the big road and its derived roads — and tells you whether to bet Banker or Player next. Follow the streak, bet the chop, break the dragon, the "1-2" and "3-pattern" rules, and the countless "secret baccarat pattern strategy" videos are all versions of the same idea: the recent shape of results predicts the next one. This page does two things. First it writes down exactly what each popular pattern bets on, so you can see they are testable rules rather than instincts. Then it reports what happened when we tested them on 118,203 completed real shoes — 8.52 million hands — from live casino tables, with the sample sizes attached. The short version: every pattern we could define ended up where an 8-deck shoe says it should, within a fraction of a percent of the house edge.
The popular baccarat pattern strategies, defined
| Pattern strategy | What you bet | The claim behind it |
|---|---|---|
| Follow the shoe (ride the streak) | Whatever won the last hand | Streaks "run" — a side that is winning keeps winning |
| Chop (ping-pong, zigzag) | The opposite of the last hand | Results alternate — a choppy shoe stays choppy |
| Break the dragon | Against a streak once it reaches 4, 5 or 6 | Long streaks are "due" to end |
| 1-2 rule | After two of the same side, switch | Pairs are more common than longer runs |
| 3-pattern | After three of the same side, stay on it | Three in a row signals a dragon is forming |
| Follow the derived roads | What the big eye, small and cockroach roads indicate (red = repeat, blue = change) | Derived roads reveal "predictability" in the shoe |
| Trend switching | Switch between follow and chop after two losses | Shoes have "modes" you can latch onto |
| Mirror / repeating shoe | Bet that the shoe repeats an earlier section | Shuffles leave structure behind |
Each of these is a precise instruction: given the last few results, bet Banker or Player. That makes every one of them measurable. If a pattern strategy works, the side it tells you to bet must win more often than that side wins in general — Banker more than 50.68% of decided hands, Player more than 49.32%. Those two numbers are the bar.
What the arithmetic says before any data
Punto banco has no memory. The cards are shuffled, the drawing rules are fixed, and the only thing linking one hand to the next is the removal of four to six cards from a 416-card shoe — far too little to move the next result measurably. So the probability of the next decided hand is always close to Banker 50.68%, Player 49.32%, whatever the scoreboard shows. From that, the basic pattern quantities follow directly:
| Quantity | Value | What it means for pattern bettors |
|---|---|---|
| Next decision repeats the last one | 50.01% | "Follow" wins half the time |
| Next decision changes | 49.99% | "Chop" wins half the time |
| Columns on the big road per 80-hand shoe | about 36.5 | Average column is 2.0 hands long |
| Column of 2 or more | 50.0% of columns | Pairs are normal |
| Column of 4 or more | 12.5% of columns | A 4-streak is 1 column in 8 |
| Column of 7 or more ("dragon tail") | 1.57% of columns | About 0.5 per shoe; 42% of shoes have one |
And the odds of seeing a long streak somewhere in a shoe are far higher than most players feel:
| Streak of at least k (same side, ties ignored) | Chance of at least one in a 72-decision shoe | Expected number per shoe |
|---|---|---|
| 4 | 99.7% | 4.38 |
| 5 | 92.1% | 2.16 |
| 6 | 68.8% | 1.06 |
| 7 | 42.4% | 0.53 |
| 8 | 23.4% | 0.26 |
| 10 | 6.2% | 0.063 |
| 12 | 1.5% | 0.015 |
A dragon of six or more appears in two shoes out of three. "Break the dragon" bettors therefore get a chance most shoes — and the hand after a 6-streak is still a 50.68/49.32 proposition, exactly like the hand after a chop.
Tested on 8.52 million real hands
Theory is one thing; the question players actually ask is whether real casino shoes behave like the theory. We keep an archive of completed shoes from live Banker/Player tables. The discovery sample was 118,203 shoes (8.52 million hands); a separate set of 12,985 shoes (937,000 hands) with the full card log was held back for validation. Results:
| Test | Result on real shoes | What the theory predicts |
|---|---|---|
| Banker streak continues for one more hand, by streak length | 50.5%–50.9% at every length | 50.68% |
| Player streak continues, by length | 49.2%–49.5% at every length | 49.32% |
| Always bet Banker (7.17 million decisions) | 50.65% hit rate | 50.68% |
| Pattern rules tested (19 pattern families × position × B/P/T) | 2,310 tests, 0 survive a multiple-comparison correction | 0 expected if there is no signal |
| Top 30 patterns from the discovery sample, re-run on the validation sample | −1.90% per hand combined | About −1.1% to −1.2% (house edge) |
| 84 app-style strategies (patterns, three-column rules, derived roads, AI model), walk-forward on 136,825 shoes | Every one between −1.0% and −1.3% per hand, hit rates 49.7%–50.7% | −1.06% (Banker) to −1.24% (Player) |
| Progressions layered on patterns (13 systems × 14 signals × sizing/targets = 3,810 settings) | 59 were positive on 6,000 shoes; 0 of 59 stayed positive on 10,000 and 120,825 fresh shoes | 0 |
The most useful row is the first. "Follow the shoe" rests on the belief that a streak that has already run four or five is more likely to continue than a fresh one. On 8.5 million hands the continuation rate did not move with length — a Banker streak of one, four or eight all continued between 50.5% and 50.9% of the time, which is the ordinary Banker win rate with sampling noise. "Break the dragon" rests on the opposite belief, and fails for the same reason. The 2,310-test row shows why the patterns in videos look convincing: test enough rules on enough past shoes and some will show 55% or 60% by chance; the ones that did here fell back to the house edge the moment they were run on shoes they had not seen.
Worked example: one shoe of "follow the shoe"
Take a flat 1-unit bettor who follows the last result through an 80-hand shoe. Ties push, leaving about 72 decided hands. The bettor is on Banker roughly half the time and on Player the other half, so the expected cost per hand sits between the two edges, 1.06% and 1.24%.
| Over one 72-decision shoe | Follow the shoe | Bet Banker every hand |
|---|---|---|
| Hands won | about 36 (50.0%) | about 36.5 (50.68%) |
| Expected result | −0.8 to −0.9 units | −0.76 units |
| Typical swing (one standard deviation) | about ±8 units | ±8.3 units |
| Chance to be ahead after 100 hands | about 45% either way | |
The pattern rule does not create wins; it only decides which of two slightly losing bets you hold on each hand. Its expected result per shoe is a little worse than plain Banker because it spends half its hands on the more expensive Player bet. Meanwhile the swing of ±8 units per shoe is ten times the expected loss, which is why a pattern "works" for a weekend — and why no weekend proves anything.
Why a pattern can look like it works for weeks
The Banker share of a single 72-decision shoe falls between 39% and 62% in 95% of shoes. Ten shoes narrow that to 47%–54%. To tell a strategy with a real 1-percentage-point edge from a coin flip you need about 133 shoes (9,600 decisions); for half a point, about 533 shoes. A pattern that "hit 62% over the last 50 hands" is sitting comfortably inside noise. When you evaluate any pattern strategy — yours, a friend's, or one sold online — ask for its hit rate over thousands of hands with the count attached. Our pattern analysis guide shows what an honest tracker computes and publishes its own measured continuation rates.
Pattern strategy plus a progression
Most pattern systems are sold together with a betting progression: double after a loss, or step 1-3-2-4, until the pattern "comes back". The progression changes the shape of your results but not their sum. Betting against a streak with a seven-step Martingale starting at 1 unit means risking 127 units to win 1, and a seven-streak against you arrives in 42.4% of shoes; the cycle busts about once in 141 attempts. Over a session that reaches a +10 unit target 91% of the time, the average result is still −1.3 units, because the 9% of sessions that fail lose around 127. The exact tables are on baccarat martingale and baccarat betting systems.
What baccarat patterns are actually good for
- Keeping score honestly. A big road tells you, at a glance, how the shoe has gone — not how it will go. Learn the notation on reading the big road and the derived roads.
- Measuring your own idea. If you believe in a pattern, write it down as a rule and track it across hundreds of shoes before betting on it. The baccarat pattern chart tool does the counting for you on live shoes, and the free Baccarat Scoreboard app keeps your own shoes offline.
- Pacing. Some players bet only on a pattern signal to play fewer hands per hour. That lowers hourly cost — not because the pattern wins, but because fewer units cross the table.
A plan that survives the numbers is short: take the smaller-edge bet, keep stakes flat and small relative to the bankroll, and decide in advance how long you play. It is spelled out in baccarat strategy.
Baccarat pattern strategy FAQ
Is there a secret baccarat pattern strategy that works? None that we could find. 2,310 pattern rules on 8.5 million hands produced zero that beat the house edge once tested on shoes they had not been fitted to.
What is the best pattern to follow in baccarat? Measured on real shoes, following, chopping and breaking streaks all land at the house edge. If you want a rule, "bet Banker" is the cheapest one: 1.06% against 1.24% for Player.
Do long streaks mean the next result is more likely to switch? No. After a Banker streak of any length the next decided hand was Banker 50.5%–50.9% of the time in 8.5 million real hands — the same as after any other hand.
Do the derived roads (big eye, small road, cockroach) predict the next hand? They describe the regularity of the past columns. Rules built on them were part of the 84 strategies tested and finished between −1.0% and −1.3% per hand.
How many hands do I need to prove a pattern works? About 133 shoes to detect a 1-point edge, 533 for half a point. Anything shorter is inside normal variance.
Why do pattern videos show such high hit rates? They are usually one short sample, a rule chosen after seeing the results, or both. Ask for the hand count and for shoes the rule had not seen.
Watch how real shoes form their patterns, and how often each one actually continues, on the live baccarat terminal.
Theoretical figures: exact enumeration of an 8-deck shoe under punto banco rules, streak figures for 72 decided hands with ties ignored. Real-shoe figures: Baccarat Terminal shoe archive, 118,203 completed shoes / 8.52 million hands (June–September 2026) with 12,985 card-logged shoes / 937,000 hands held out for validation; 84-strategy walk-forward on 136,825 shoes (September 2026). Past results do not predict future results. 18+. Information only — not a gambling service and not betting advice.