Edge Sorting in Baccarat: How One Tiny Asymmetry Beat the Casino

The casino had every structural advantage in the world — a roughly 1.06% edge on Banker, 1.24% on Player, an automatic shuffler, surveillance cameras, and a multi-deck shoe. And it still managed to lose roughly $21 million across two properties in 2012, because a professional gambler and his partner figured out that the diamonds on the back of certain playing cards weren’t quite symmetrical. That’s the entire story of edge sorting, and it’s wild that it took the gaming industry decades to take it seriously.

What Edge Sorting Actually Is

Most playing cards have a decorative pattern on the back — diamonds, lattices, swirls. You’d think the pattern is identical on all four edges of every card. It often isn’t. On certain printer runs, the pattern is trimmed slightly off-center, so one long edge shows a complete row of diamonds and the other shows half-diamonds. The asymmetry is small, often under a millimeter, but it’s consistent across the whole deck.

That tiny printing defect doesn’t matter in any normal game. Cards get rotated randomly, shuffled, and dealt from a shoe — the orientation washes out. But if you can convince the dealer to rotate certain cards 180° before they go back into the shuffler, and that shuffler doesn’t re-randomize orientation, you’ve quietly sorted the deck into two groups: cards you care about, and cards you don’t. From the back. Without ever touching them.

In baccarat, that’s enormous. Knowing whether the next card is “high” (6, 7, 8, 9) or “low” (0, 1, 2, 3, 4, 5) lets you predict whether Player or Banker is going to win that hand often enough to flip the house edge into a player edge — and not a thin one.

The Phil Ivey Case in Two Paragraphs

In 2012, Phil Ivey — one of the most accomplished poker players alive — walked into Crockfords in London with Cheung Yin Sun, a player who’d already developed edge-sorting techniques. They asked for purple Gemaco decks, a specific automatic shuffler that didn’t rotate cards, a Mandarin-speaking dealer, and the same shoe used over multiple sessions. They also asked the dealer to rotate certain cards 180° because they were “lucky.” The dealer obliged. Crockfords took the hit — about £7.7 million (roughly $11M at the time) — and then refused to pay, claiming Ivey had cheated.

A few months later, the same duo did it again at Borgata in Atlantic City for about $9.6 million, mostly at high-stakes punto banco. Borgata paid, then sued. Both cases ended badly for Ivey. The UK Supreme Court ruled against him in 2017. So did the US 3rd Circuit. The courts agreed: even though Ivey never touched a card, the requests he made compromised the integrity of the game, and that was enough to call it cheating under contract law. I’ll get to my own opinion on that further down, because the legal reasoning is genuinely interesting and not as obvious as people pretend.

The Math: Why One Card’s Identity Changes Everything

Baccarat is one of the most rules-rigid games in the casino. The player and banker each get two cards; whether a third is drawn follows a fixed table. Nobody decides anything. Because the rules are fixed, the value of any specific information about the next card is exactly computable.

The Wizard of Odds did the full calculation (wizardofodds.com/games/baccarat/edge-sorting/), and the result is striking. If you know the value of just the first card dealt to the Player — not the rank, just whether it’s a “high” card (6-9) or “low” card (0-5) — you can choose which side to bet, and your edge flips dramatically. Here’s the relevant table, simplified.

Notice what’s happening. A first card of 8 or 9 is essentially a natural in disguise — Player can’t draw a third card on a two-card 8 or 9, and a high first card makes a natural very likely. A first card of 0-3 is the opposite story: Banker is heavily favored because Player will probably need to draw and end up with a mediocre hand. The middle cards (4, 5, 6, 7) are softer edges, but they’re still edges.

Average it out across the distribution of “high” vs. “low” first cards and you land somewhere around a +6.7% advantage to the sorter who plays correctly. That’s not blackjack card-counting territory — that’s a number that makes the house feel like the underdog.

What Had to Go Right for Edge Sorting to Work

The setup is what makes this fascinating. Knowing the asymmetry exists isn’t enough. You need a chain of conditions, and any single broken link kills the whole thing.

• A flawed deck. The classic case used purple Gemaco Borgata decks, where the diamond border was trimmed asymmetrically. Not every deck has this — it’s a manufacturing defect, not a feature.
• A cooperative dealer. Someone has to rotate the “lucky” cards 180° as they’re scooped up. In a high-roller pit, players can ask for almost anything, and dealers will accommodate it. A pit boss watching a $100,000-a-hand whale isn’t going to argue about superstition.
• A non-rotating shuffler. Automatic shufflers usually preserve orientation. A human dealer doing a riffle shuffle would scramble the rotation and erase the sort instantly.
• The same shoe over multiple sessions. The sort accumulates. One night isn’t enough to fully separate the high cards from the low cards — you need the decks to come back the next day in the same orientation.
• A sharp eye on the back of the top card. You’re reading a millimeter-wide pattern asymmetry under casino lighting in real time. This isn’t a casual skill.

Strip out the shuffler. Bring in a hand shuffle. Use a fresh deck every shoe. Any one of those, and the math collapses back to the baseline house edge. The casinos didn’t lose because the technique was unstoppable — they lost because they accommodated every single one of those requests without thinking about why someone was asking.

Why the Casinos Should’ve Caught On Sooner

This is the part that genuinely surprises me. The math has been knowable for decades. The Gemaco asymmetry wasn’t a state secret — printers have been making slightly off-center cards forever. And the rules of baccarat are so brittle that anyone who’s thought about the game seriously knows the third-card draws hinge on small information.

What protected the industry wasn’t security. It was the assumption that no player would care enough to combine all five conditions above into a single coordinated attempt. Ivey and Sun were essentially exploiting institutional laziness — the polite, white-glove version of high-roller hospitality where any request gets a “yes.”

If you’ve ever taught yourself a topic where the textbook makes the underlying logic feel a lot harder than it is — and a quick walk through the structure makes the whole thing click — you’ll recognize the pattern. (Side note: that’s roughly the philosophy behind Effortless Math — small structural insights tend to do more work than memorization.) The casinos weren’t undone by a complicated trick. They were undone by one structural truth they’d never bothered to map: the back of a card carries information when its orientation is preserved.

The Legal Side: Was It Cheating?

Here’s where I’ll be honest about my own take. I don’t think the Ivey case is morally clean in either direction.

The pro-Ivey argument is straightforward: he never touched a card, never marked anything, never colluded with a dealer to alter the game. He observed an existing physical property of the deck and made strategic requests that the casino was free to refuse. Casinos throw out card counters all the time without paying them — they don’t sue them for the winnings. Ivey just played better.

The pro-casino argument is also strong, though. By asking the dealer to rotate certain cards, Ivey was changing the physical state of the equipment the game runs on. He wasn’t just using public information — he was actively engineering the information’s existence through dealer requests made under the cover of superstition. That’s a different thing from card counting, where the count exists whether you ask for anything or not.

The courts ultimately ran with the second framing. In both jurisdictions, the rulings centered on the dishonest representation that the card rotation was a superstition rather than an exploitation. The legal concept lives roughly in the same neighborhood as fraud-by-deception, not fraud-by-action. If you want the dry version, the Cornell Legal Information Institute entry on edge sorting is short and worth reading.

My honest read? Crockfords and Borgata were sloppy, and Ivey’s team was clever, but the deception-by-superstition piece is what tips it from “advantage play” to something the courts can defensibly call cheating. If he’d simply observed the asymmetry across a fairly-shuffled shoe and bet accordingly, I’d be on his side. The dealer-rotation request is what costs him the moral high ground for me.

What It Means for the Rest of Us

You’re not going to edge-sort a baccarat shoe. The flawed Gemaco decks have been pulled, casinos rotate cards before they hit the shuffler, and any request for a specific deck or dealer behavior now gets logged and flagged in seconds. The window closed in 2013-2014.

What’s left is the lesson. The Ivey case is one of the cleanest real-world examples of how a tiny, ignored asymmetry can dominate a large, well-designed system. The house edge in baccarat is roughly 1%. The exploit pushed the edge to roughly 7% — about a 7-to-1 swing — driven by a printing defect under a millimeter wide. That’s the kind of leverage that mathematics can give an observer, and it’s the reason every serious gambling operation now treats card asymmetry as a manufacturing-grade specification, not a cosmetic detail.

FAQ

Did Phil Ivey ever get to keep any of the money?
No. Crockfords refused to pay from the start. Borgata paid initially but sued and won back its $10.1M, plus additional damages on appeal. Ivey ended up owing rather than collecting.

Are Gemaco cards still used in casinos?
Gemaco still supplies cards to many casinos, but the specific flawed runs implicated in the Ivey cases were pulled. Modern quality control on casino decks is much tighter — manufacturers now spec the back-pattern symmetry as a tolerance, not just a print preference.

Could edge sorting work in blackjack?
In principle, yes — and there are reports it has been tried — but blackjack involves player decisions, which dilutes the value of partial information. The fixed-rule structure of baccarat is what made the technique so devastating there.

Is just observing an asymmetric back legal?
This is the unsettled part. Pure observation of a physical defect, without asking the dealer to do anything unusual, hasn’t been tested in court as far as I know. The Ivey cases turned on the dealer-rotation request, not the observation itself. A pure observer might have a real defense.

What’s the actual edge percentage for a perfect edge sorter?
Roughly +6.7% on Player bets under the conditions described, though it ranges depending on how completely the shoe is sorted and how disciplined the bettor is about switching sides based on the first card. Wizard of Odds is the standard reference for the full hand-by-hand calculation.

Closing Thought

Edge sorting is, in a strange way, a story about taking small numbers seriously. A one-millimeter trim error sounds like nothing. A 1.06% house edge sounds like nothing. Combine them with a non-rotating shuffler and a dealer who’ll accommodate a request, and suddenly two people are walking out of casinos with sums of money that look like rounding errors on a Fortune 500 balance sheet — except they aren’t, because someone added them up correctly.

I find the case more interesting than scandalous. Two people noticed something the casino didn’t, and the casino’s response — sue rather than upgrade your equipment — felt like the wrong instinct. The deck didn’t get fixed because of the lawsuit. It got fixed because the industry finally measured what it had been assuming. Which is usually how these stories end, in any field that takes math seriously.