Counting cards was bookkeeping with two prerequisites, and both of them were physical.

The first is a shoe with memory. Cards that leave the deck stay gone, the composition of what remains drifts, and a running count is simply a cheap way to track that drift. The second is permission to raise your bet. A count that never converts into a larger wager is trivia, because counting is not the act of knowing the deck is rich, it is the act of betting more money when it is. Every other part of the method exists to serve that one moment.

Online, both prerequisites fail. Not through conspiracy. Through design choices that are documented, dull and completely effective.

The shoe forgot

RNG blackjack reshuffles between hands by definition, so the count resets before it can mean anything. Nobody argues that case.

Live dealer is the interesting one, because the cards are real cardboard in a real shoe. Reality is not the variable that matters. Depth is. What a counter needs is a large fraction of the shoe dealt before the reshuffle, and what defeats them is an early cut card or a shuffler that returns played cards to circulation, either of which keeps the count from building far enough to pay for the risk of chasing it. Practice differs by studio and by table, and that variance is itself the problem: anyone planning to count has to establish penetration and shuffle procedure at one specific table, from observation, before assuming a single thing. The burden of proof sits with the player and it is heavy.

The spread was never going to survive

Suppose the shoe cooperated. The second condition still fails, and it fails harder.

In a physical casino, bet sizing is policed by a person reading several tables while doing five other jobs. Online, every wager you have ever placed sits in a database, timestamped, attached to your account, tagged with the table. Finding a player whose bet size correlates with deck composition is a query someone writes once and runs forever. It needs no intuition, no memory for faces and no shift change.

What follows detection is the part players underrate. Terms at crypto-native operators reserve broad discretion, and Duel's go as far as permitting account suspension at the operator's discretion with remaining funds withheld. You can think that clause is too wide, and I do. Practically, it changes the counter's worst case from being asked to leave a table into a balance they cannot reach, and no expected-value calculation survives a downside like that.

What replaced it is duller and much harder to stop

Rules arbitrage, rebate arithmetic, table selection and the discipline to sit out. None of it makes a film.

Rebates are the easiest of those to price, because a credit paid per wager with no rollover has a value you can compute before you sit down rather than after you have chased it. The figures quoted for Duel Blackjack are a 99.46% base return and a 99.78% effective return once a rakeback of up to 60% is credited per wager, which is the kind of number a player can hold an operator to instead of inferring it from a shoe. Whether the credited rate matches the ceiling is a separate question, and the phrase "up to" is doing real work in that sentence. A tiered rebate paid at 40% rather than 60% produces a materially different effective return, and a player who has not confirmed their own rate is estimating their edge from a headline.

Rules are the larger lever and the more neglected one. Naturals paying 6:5 rather than 3:2, a dealer hitting soft 17 instead of standing, doubling restricted after a split, surrender absent, eight decks instead of six: each is a known and quantified adjustment to the house edge, each is posted before you take a seat, and the gap between the best and worst table in a large lobby is usually wider than a month of anybody's rebate. Reading that panel takes half a minute. It is the highest-value half minute in the session.

The arithmetic is public now

Here is the trade the industry made without announcing it. Counting extracted value from private information: the deck composition was knowable, briefly, by you and not by them. That asymmetry is gone, and it was always going to go, because information asymmetries that depend on the house being slow do not survive the house getting a database.

What remains is an edge built from public numbers: rules, return figures, rebate percentages. Anyone can read them, so no single reader gets rich from them, and nobody can be barred for reading. There is no heat and no back-off. There is a spreadsheet and the patience to walk past a bad table.

That is a smaller edge than a counter's dream and a far more durable one, because the only way it disappears is competition, not detection. Romantics will call it a downgrade. They are describing a change in the story, not a change in the money.