Baccarat Roadmaps, Hidden State, and the Mathematics of Misread Patterns

Abstract

Baccarat roadmaps are among the most familiar visual systems in modern casino gaming. They are also among the most misunderstood. The Big Road, Big Eye Boy, Small Road, and Cockroach Pig appear to give order to the shoe: streaks, chops, repeated structures, broken structures, and apparent rhythms. Yet the mathematical object that determines the next hand is not the pattern on the display. It is the remaining composition of the shoe.

This paper argues that baccarat roadmaps fail as predictive tools because they record terminal outcomes rather than measure the hidden card state. Baccarat is not perfectly independent in the strict mathematical sense—cards are dealt without replacement—but the information available to a roadmap player is too compressed to recover the shoe composition with useful precision. The derived roads deepen the illusion: they transform the Big Road into secondary visual forms without adding new information about the undealt cards.

Three interlocking claims follow. First, baccarat is modeled as a finite-shoe stochastic process whose true state is the vector of remaining card values. Second, the derived roads are deterministic transformations of the Big Road, and therefore incapable of increasing information about the hidden shoe state. Third, a Monte Carlo simulation of 50,000 eight-deck shoes tests common roadmap betting rules against the standard Banker and Player wagers. The results are consistent with exact baccarat probabilities: all tested roadmap systems remain negative expectation.

What follows is not simply an argument that patterns do not work. The stronger claim is that baccarat roadmaps describe the past in a visually persuasive form while leaving the variable that actually governs future expectation unobserved.

I. Introduction: The Screen Looks Smarter Than the Game Allows

A baccarat table rarely feels like a bare probability experiment. The game is surrounded by ritual: the slow squeeze of the cards, the silence before a reveal, the language of “Banker run,” “Player comeback,” “dragon,” “chop,” and “broken road.” Above or beside the table sits the display that gives these ideas a visual body. The Bead Plate records outcomes in sequence. The Big Road rearranges them into streaks and changes. The Big Eye Boy, Small Road, and Cockroach Pig go further, marking whether the Big Road is behaving regularly or irregularly.

The rules of the underlying game are simple. In the usual game, eight decks are used; aces count as 1, cards 2–9 count as their pip value, and tens and face cards count as 0. The Player and Banker hands receive two cards each, and third-card decisions are governed by a fixed tableau rather than by player choice. Banker wins pay 19 to 20 because of the 5% commission, while Player wins pay even money. (Wizard of Odds)

The display, however, makes the game feel less simple. It invites interpretation. A player does not merely ask, “Which bet has the lowest house edge?” He asks whether the shoe is running, chopping, repeating, or breaking. At that point the analysis quietly goes wrong: the roadmap is treated as a diagnostic instrument when it is only a historical drawing.

The exact eight-deck analysis gives the Banker bet a return of (−0.010579), the Player bet a return of (−0.012351), and the Tie bet a return of (−0.143596)—house edges of approximately 1.06%, 1.24%, and 14.36% respectively. (Wizard of Odds) Those numbers are not changed by a beautiful Big Road.

II. The True State of the Game

Let the state of an eight-deck baccarat shoe before hand (n) be written as

[ C_n=(c_0,c_1,c_2,\ldots,c_9), ]

where (c_0) is the number of remaining zero-value cards and (c_i), for (i=1,\ldots,9), is the number of remaining cards with baccarat value (i). At the beginning of an eight-deck shoe there are 416 cards. As hands are dealt, the vector changes.

The next outcome is

[ X_{n+1}\in{B,P,T}, ]

where (B) means Banker, (P) means Player, and (T) means Tie.

Because baccarat is dealt without replacement, the exact probability of the next hand depends on (C_n):

[ P(X_{n+1}=x\mid C_n). ]

Baccarat is not literally identical to roulette—the shoe does change. But that observation does not rescue the roadmap player. The roadmap does not show (C_n). It shows only previous outcomes:

[ R_n=(X_1,X_2,\ldots,X_n). ]

The difference between (C_n) and (R_n) is the difference between the real state of the shoe and the public story of the shoe. The real state contains the remaining cards. The public story contains only Banker, Player, and Tie labels.

For a unit Banker wager under the standard commission rule, the conditional expected value is

[ EV_B(C_n)=0.95P(B\mid C_n)-P(P\mid C_n). ]

For a unit Player wager,

[ EV_P(C_n)=P(P\mid C_n)-P(B\mid C_n). ]

A profitable baccarat decision would require identifying states where one of these quantities becomes positive. Roadmaps do not measure those states directly.

III. Proof One: The Roadmap Is Not the Shoe

Proposition 1

The visible baccarat roadmap (R_n) is not a sufficient statistic for the shoe composition (C_n).

Proof

A roadmap records the terminal outcome of each completed hand: Banker, Player, or Tie. It does not record the exact cards dealt, their order, the third-card path, or the card values removed from the shoe.

Many different card sequences can produce the same visible roadmap. A Banker win can arise through a large number of card combinations. Some of those combinations remove high cards; others remove low cards; some deplete zero-value cards heavily while others do not. Yet the roadmap records all of them as the same symbol: Banker.

Therefore there exist distinct shoe states (C_n^{(1)}) and (C_n^{(2)}) such that

[ R_n(C_n^{(1)})=R_n(C_n^{(2)}), ]

but

[ C_n^{(1)}\neq C_n^{(2)}. ]

If two different shoe compositions can produce the same roadmap, the road cannot identify the shoe composition. The player is predicting a composition-dependent process from a display that does not preserve composition.

IV. The Derived Roads: Sophistication Without New Information

The derived roads are often treated as more advanced than the Big Road. That belief is understandable—they are harder to read, and their red and blue marks do not directly mean Banker and Player. They seem to operate one level deeper.

In the mathematical sense, however, they are not deeper at all. They are derived from the Big Road itself. Contemporary baccarat-road descriptions explain the Big Eye Boy, Small Road, and Cockroach Pig as roads that track regularity in the Big Road structure by comparing the current column with earlier columns at increasing offsets—not by observing the cards. (Mayfair Casino London)

Let the Big Road be

[ R_n^{(0)}=f_0(X_1,\ldots,X_n). ]

Let the derived roads be

[ R_n^{(1)}=f_1(R_n^{(0)}), ]

[ R_n^{(2)}=f_2(R_n^{(0)}), ]

[ R_n^{(3)}=f_3(R_n^{(0)}). ]

Here (R_n^{(1)}) is the Big Eye Boy, (R_n^{(2)}) is the Small Road, and (R_n^{(3)}) is the Cockroach Pig.

The notation matters. The derived roads are functions of the Big Road. They are not functions of the undealt cards.

V. Proof Two: Derived Roads Cannot Add Hidden-State Information

Proposition 2

A deterministic transformation of the Big Road cannot contain more information about the shoe composition than the Big Road itself.

Proof

Each derived road is produced by applying a fixed rule to the Big Road. Thus,

[ R_n^{(i)}=f_i(R_n^{(0)}),\quad i=1,2,3. ]

In information-theoretic terms, each derived road is post-processing of an already-compressed signal. It does not observe the shoe, does not observe the removed cards, and does not observe the next card. It only rearranges or reclassifies information already present in (R_n^{(0)}).

By the data-processing inequality, deterministic post-processing cannot increase information about an unobserved variable. Therefore,

[ I(C_n;R_n^{(1)},R_n^{(2)},R_n^{(3)})\leq I(C_n;R_n^{(0)}). ]

If the Big Road does not know the undealt shoe composition, its children do not know it either. The derived roads may change the appearance of the evidence; they do not change its informational content. They classify the shape of the past.

VI. Effect of Removal: The Strongest Objection

An obvious counterargument: baccarat cards are dealt without replacement, so past hands must in some degree affect future ones.

The objection is well-founded as far as it goes. Card removal does matter. The question is whether the roadmap captures enough of that removal to create a betting edge.

Effect of Removal, or EoR, measures how the removal of a specific card changes the house edge or win probability. In blackjack, EoR is large enough to support card counting under favorable rules and deep penetration. Baccarat is different. The Wizard of Odds effect-of-removal table for eight-deck baccarat shows that after one card is removed, Banker-win probability remains in a narrow band, roughly from 0.458534 to 0.458673 depending on the removed card value; Player and Tie probabilities also move only slightly. (Wizard of Odds)

The corresponding Banker and Player house edges after removing one card are real, but very small. (Wizard of Odds) The same analysis reports extremely high true-count thresholds for reaching zero house edge: 105,791 for Banker, 123,508 for Player, and 1,435,963 for Tie, with positive-expectation opportunities appearing rarely even under perfect-count assumptions and deep penetration. (Wizard of Odds)

Thorp and Walden’s 1973 paper treated sampling without replacement as a general framework for card games, applying it formally to trente-et-quarante and baccarat. Their analysis confirms that composition effects can be quantified—and that the thresholds required for practical advantage are far beyond what visual history can detect. (Edward O. Thorp)

The shoe composition is relevant. The roadmap is too crude to recover it:

[ P(X_{n+1}\mid C_n)\neq P(X_{n+1}), ]

but for ordinary roadmap play,

[ P(X_{n+1}\mid R_n)\approx P(X_{n+1}). ]

VII. Proof Three: What a Roadmap Strategy Would Have to Prove

Let a roadmap strategy be a rule

[ S_n=g(R_n), ]

where (S_n\in{B,P,\varnothing}). The symbol (\varnothing) means no bet.

The expected return of the strategy is

[ EV(S)=E[\pi(S_n,X_{n+1})], ]

where (\pi) is the payoff function.

For the strategy to be genuinely advantageous, it must satisfy

[ EV(S)>0. ]

That holds only if the roadmap rule selects a subset of hands where the conditional payoff is positive:

[ E[\pi(S_n,X_{n+1})\mid S_n\neq \varnothing]>0. ]

This is an empirical claim, not an impossibility. But it cannot be established by pointing to a red streak, a repeated column shape, or a “strong” Cockroach Pig signal. Any roadmap system claiming advantage must show that its visual trigger identifies positive-expectation shoe states. Without that demonstration, the rule is a method for choosing among negative-expectation wagers—nothing more.

The conclusion preserves rather than dismisses the card-composition argument; it only notes that visual roadmaps are not a practical instrument for exploiting it.

[ \boxed{\text{A roadmap signal is not evidence of advantage unless it predicts conditional expected value.}} ]

VIII. Monte Carlo Test

To test representative roadmap ideas, I ran a Monte Carlo simulation of standard eight-deck Punto Banco.

Simulation parameters

The simulation used the standard baccarat dealing rules: two cards to Player and Banker, natural 8 or 9 standing rule, Player drawing on 0–5, and the Banker third-card tableau. These are the same structural rules summarized in the Wizard of Odds baccarat rules. (Wizard of Odds)

Observed outcome frequencies

These frequencies are close to the exact eight-deck probabilities: Banker 0.458597, Player 0.446247, Tie 0.095156. (Wizard of Odds)

Tested betting rules

The simulation tested six common classes of baccarat betting behavior:

These rules do not exhaust every possible roadmap system. They test the most common claims: follow the run, oppose the run, follow the chop, or wait for a long enough streak before acting.

Simulation results

No tested rule produced positive expectation. The returns cluster around the known Banker and Player disadvantage. The streak systems did not become stronger by waiting for longer runs. Trend and chop rules found no exploitable serial dependence—they simply redistributed betting volume across the same negative-expectation environment.

These results do not rule out every conceivable history-based rule; that would be an overclaim. What they show is more specific: the roadmap heuristics most familiar to players do not overcome the underlying expectation, and any claimed exception should be reproducible at the same scale.

IX. Why the Roads Persuade the Eye

Random sequences do not look smooth in short samples. They cluster, alternate, pause, and run. A player who expects randomness to appear evenly balanced will read ordinary clustering as evidence of structure, and baccarat roadmaps provide exactly the visual grammar that makes such readings feel rigorous.

Tversky and Kahneman identified what they called the “law of small numbers”—the tendency to expect small samples to resemble the parent distribution more closely than they actually do—and showed that people systematically overestimate how much short sequences reveal about underlying processes. (CiNii Research) Gilovich, Vallone, and Tversky later connected streak perception to the hot-hand fallacy, finding that observers attribute genuine momentum to sequences where the evidence supports no such inference. (ScienceDirect) The sports-specific hot-hand debate has grown more complicated since then, but baccarat is a cleaner case: the Player and Banker hands do not become “hot.” The cards are dealt by rule from a finite shoe. The display does not feed back into it.

Roadmaps work on the imagination precisely because they turn variance into landscape. A long Banker column becomes a “dragon.” Alternation becomes a “chop.” A red derived road signals “order.” A blue one signals “chaos.” None of these labels appears in the game’s dealing rules.

X. Casino Operations: Why the Display Remains Useful

If roadmaps do not beat the game, why do casinos provide them?

A tool does not need to improve player expectation to improve player engagement. Roadmaps reduce hesitation. They give players something to study between hands. They create a sense of participation in a game where no strategic decision remains once the wager is placed.

A blank table asks the player to accept randomness. A roadmap gives the player a story. The story may be false as prediction, but it is effective as experience.

From an operations standpoint, that is a distinction worth holding clearly. The screen is not cheating; it is recording. But the form of the record encourages a belief the mathematics does not support.

XI. Conclusion

Baccarat roadmaps fail because they do not observe the state that determines future expectation.

The true state of a baccarat shoe is the remaining composition vector (C_n). The roadmap (R_n) is a compressed history of terminal outcomes. Since many different card-removal paths can produce the same visible road, the road cannot identify the shoe composition. The derived roads are weaker still as independent evidence: deterministic transformations of the Big Road, they may alter the visual grammar of the past but add nothing about the undealt cards.

The exact eight-deck game remains negative expectation: approximately -1.06% on Banker, -1.24% on Player, and -14.36% on Tie. (Wizard of Odds) Effect-of-removal data confirms that card depletion has real but small effects, and baccarat card-counting thresholds are so extreme that visual roadmap play offers no practical route to advantage. (Wizard of Odds)

The simulation reinforces the theory. Across 50,000 shoes and more than four million hands, trend-following, chop-following, streak-continuation, and streak-reversal rules all remained negative. They changed when the player bet, but not the fundamental expectation of the wager.

[ \boxed{ \text{Baccarat roadmaps are excellent records of what happened and poor instruments for knowing what will happen next.} } ]

They do not reveal the shoe. They decorate its history.

Bibliography

Ethier, Stewart N., and Jiyeon Lee. “The Evolution of the Game of Baccarat.” arXiv, 2013. The paper traces baccarat’s movement from strategic parlor forms toward the modern nonstrategic casino game. (arXiv)

Gilovich, Thomas, Robert Vallone, and Amos Tversky. “The Hot Hand in Basketball: On the Misperception of Random Sequences.” Cognitive Psychology 17, no. 3 (1985): 295–314. DOI: 10.1016/0010-0285(85)90010-6. (ScienceDirect)

Mayfair Casino. “Reading the Baccarat Roads.” Reviewed May 14, 2026. Used for contemporary descriptions of the Big Road, Big Eye Boy, Small Road, and Cockroach Pig. (Mayfair Casino London)

Thorp, Edward O., and William E. Walden. “The Fundamental Theorem of Card Counting with Applications to Trente-et-Quarante and Baccarat.” International Journal of Game Theory 2, no. 1 (1973): 109–119. DOI: 10.1007/BF01737563. (Edward O. Thorp)

Tversky, Amos, and Daniel Kahneman. “Belief in the Law of Small Numbers.” Psychological Bulletin 76, no. 2 (1971): 105–110. DOI: 10.1037/h0031322. (CiNii Research)

Wizard of Odds. “Baccarat.” Last updated November 11, 2024. Used for standard baccarat rules, drawing rules, eight-deck probabilities, and house-edge tables. (Wizard of Odds)

Wizard of Odds. “Baccarat Card Counting — Effects of Removing a Card.” Used for effect-of-removal probabilities, house-edge shifts, and card-counting threshold figures. (Wizard of Odds)