The short answer: a side bet is worth attacking when its payout depends on cards that can be tracked, and when the base house edge is small enough to be overcome before the shoe runs out. Most side bets fail one of those tests. The handful that pass both have a consistent history: they get attacked, they lose money, and eventually they get pulled from floors. Knowing which category a bet sits in tells a surveillance team what to watch and, just as importantly, what a suspicious betting pattern actually means.
Why the main game is not the problem
Baccarat was considered uncountable for decades, and for the main game that is essentially right. The banker and player bets depend on the overall composition of the shoe only weakly. Thorp and Griffin both worked the numbers decades ago and found the edge moves so little that no practical betting strategy emerges. A counter who tracked the whole shoe perfectly would still be waiting a very long time for a meaningful advantage.
Blackjack is the opposite case, which is why it defined the popular idea of card counting. The player's edge there is close to linear in the ratio of high cards to low cards remaining. That single fact is what makes counting tractable:
- Every card gets a fixed tag: a ten is always −1, a five is always +1, regardless of what came before
- The counter maintains one running number
- Dividing by decks remaining converts it into a direct estimate of the edge
- The house edge to overcome is roughly half a percent, and favourable shoes arrive perhaps 30% of the time
One number, a small hurdle, and frequent opportunities. That combination is rare.
What makes a side bet vulnerable
Side bets break the pattern because their payouts are tied to specific cards rather than to a general high–low balance. Two properties decide how exposed a given bet is.
How many things must be tracked. Some side bets depend on a single narrow quantity. Insurance in blackjack pays on the dealer holding a ten-value card, so it is driven by the ten-to-non-ten ratio. That is one dimension, and one that a standard blackjack count already produces for free. That is why insurance is the most exploited bet in casino history: every counter at the table is already computing the number needed to beat it.
Rank-concentration bets sit at the other extreme. A pair bet pays when two cards match, so its probability is a sum of squares across the thirteen rank counts, not a simple ratio. There is no fixed tag that works: removing a king raises the concentration in one shoe state and lowers it in another, so the correct tag depends on the composition you are trying to estimate. No single running count can represent a sum of squares. That is a mathematical limit, not a gap waiting for a cleverer system.
How big the hurdle is. A bet starting at a 2–3% house edge can flip on a moderate shift in composition. A bet starting at 12% needs an extreme one that may occur in well under 1% of shoes. The high base edge that makes a side bet profitable is also its own protection.
The bets with documented exposure
Approximate figures, because every one of these varies with paytable, deck count and jurisdiction. The numbers below are the commonly published values for standard eight-deck versions and should never be assumed for a specific table without checking its own paytable.
| Side bet | Game | Approx. base house edge | Why it is exposed |
|---|---|---|---|
| Insurance | Blackjack | ~7.5% | Driven by one ratio the counter already tracks. Universally available. |
| Dragon 7 | EZ Baccarat | ~7.6% | Wins only on a three-card banker seven; published counting systems and a history of team play. |
| Panda 8 | EZ Baccarat | ~10% | Attacked alongside Dragon 7 by the same approach. |
| Dragon Bonus | Baccarat | ~2.7% (player side; banker ~9.4%) | Low base edge on the player side means a small compositional shift is enough. Very widely available. |
| Lucky Lucky | Blackjack | ~2–7% | Low hurdle on the better paytables, and correlates with an ordinary high–low count. |
| Perfect Pairs / Any Pair | Both | ~4–11% | Rank concentration: no single count captures it, whatever the paytable. |
| Super 6 / Lucky 6 / Tiger | Baccarat | ~7–31% | Usually a large cushion, but the spread across published paytables is enormous, and the most generous variants land in Dragon 7 territory. Read the paytable on the table, not the family name. |
The pattern in that table is the useful part. The genuinely dangerous bets are not the ones with the most exotic payouts. They are the ones whose edge moves on something a person can already hold in their head.
The ones that did not survive
Several side bets were removed from most floors worldwide because they proved too easy to attack:
- Over/Under 13: one of the first side bets publicly demonstrated to be beatable, with a very high correlation to an ordinary count
- Super Sevens: extreme sensitivity to the remaining sevens, and a trivially simple tracking system
- Royal Match: suited high-card tracking, straightforward to exploit
None of these were withdrawn because someone predicted the problem. They were withdrawn after the losses arrived. The published analysis in each case came from the advantage-play side first. Eliot Jacobson's and James Grosjean's work catalogued vulnerabilities that operators were not measuring, and the games disappeared afterwards.
That is the practical argument for measuring exposure rather than reacting to it. The bet on your floor that will be withdrawn in three years is already on the floor now.
What a betting pattern actually tells you
This is where the distinction earns its keep operationally.
If a blackjack player raises their bets in step with a high–low count, the explanation is probably skilled play. It is legal in most jurisdictions, and the response is procedural: shuffle up, flat-bet them, or ask them to stop playing that game.
If a player's wagers on a rank-concentration side bet track the edge over a sustained run, that explanation is not available. The information required cannot be held in one running count, and while team approaches have been analysed, they demand several trained people, covert signalling, a very large bet spread, and long unproductive waits for opportunities that appear in a small fraction of shoes. As a business, it does not work.
So a persistent, statistically significant relationship between wager size and composition-derived edge on that class of bet points to information arriving from outside the player's head. That is a different category of problem with a different response path: escalation rather than a procedural countermeasure.
The reasoning has to be run per bet, because the answer changes with the bet's mathematical structure. A correlation on insurance means something quite different from the same correlation on a pair bet.
What this means for surveillance
Three practical consequences.
Know which bets on your floor are which. The vulnerability of a side bet is a property of its paytable and the game it sits on, and it can be worked out in advance. This is a documentation exercise, not a detection one.
Watch the wager, not just the outcome. A player who loses overall can still be extracting value, and a player on a heater is usually just lucky. The informative signal is whether bet sizing moves with the edge, a relationship that persists across a session and is not explained by the player's results.
Decide the response before you need it. Whether a finding leads to a back-off or a regulator conversation depends on which bet it is. Working that out during an incident is the wrong time.
EdgeTrack Lab tracks the true odds of a composition-dependent side bet from what remains in the shoe and correlates that edge against each player's wagers, coup by coup, from the logs your surveillance team already keeps, with no hardware and no integration project. If that is the exposure you are trying to measure, book a demo.
Sources
All figures and historical claims here come from published sources:
- Wizard of Odds: baccarat side bets and card counting the Dragon bet, for house-edge values and counting analysis
- Eliot Jacobson, Advanced Advantage Play: Beating and Safeguarding Modern Casino Table Games, Side Bets and Promotions, for side-bet vulnerability analysis from the game-protection side
- Edward O. Thorp, Beat the Dealer, and Peter Griffin, The Theory of Blackjack, for the linearity result underlying blackjack counting and the early baccarat analysis