The deck has a memory only when cards are removed
A shuffled shoe of cards changes as hands are dealt. In blackjack, the cards you have already seen are gone from the shoe until the next shuffle. That removal changes the proportion of high cards, low cards and middle cards left to be played, and that proportion changes the expected value of the next hand. Most casino games reset their odds on every round. A roulette wheel does not remember the last number. A pair of dice does not remember the last roll. A baccarat shoe does remove cards, but the structure of the game gives the player almost nothing useful to do with that information. Blackjack is different because the player makes decisions after seeing some of the cards, and those decisions can be changed to exploit what the remaining shoe contains.
The question underneath the query is really about information: which games produce information that a player can act on, and which games produce information that changes nothing. Card counting is not a trick that works everywhere but is only allowed in blackjack. It is a specific response to a specific structure, and that structure exists in almost no other table game.
What a counter is tracking
A card counter is not memorising every card. The practical systems assign point values to groups of cards, then maintain a running total as cards appear. The point of the total is to estimate whether the remaining shoe is richer or poorer in the cards that favour the player. In blackjack, high cards favour the player more than the dealer for two reasons. A ten-value card or an ace makes a natural blackjack possible, and a natural blackjack pays more than an ordinary win under the standard rule set. High cards also make the dealer more likely to bust on stiff totals, because the dealer must hit totals below a fixed threshold while the player may stand.
The counter changes two things when the remaining shoe is rich in high cards: the size of the bet and, to a lesser extent, the playing decision. The bet rises because the expected value of the next hand has shifted. The playing decision changes because basic strategy is calculated for a full or balanced deck, and a deck with an unusual proportion of high cards can make a normally marginal play slightly better or worse. Both adjustments are small, which is why counting requires many hands and a large sample before the edge becomes visible.
Why the same idea fails in roulette and craps
Roulette and craps are independent-trial games. Each spin and each roll starts from the same physical equipment and the same probability distribution. A European roulette wheel has 37 pockets and a single zero. An American wheel has 38 with a double zero. Those counts do not change because a number appeared on the last spin. The wheel has no state that the player can observe and exploit. A streak of red numbers does not make black more likely on the next spin, and no betting pattern changes the built-in edge on any single-number or outside bet.
Craps has the same property. Two six-sided dice produce 36 equally likely combinations, and the dice do not wear in a way that a player at the rail can measure from one roll to the next. Different bet types on the same table carry materially different house edges, but those edges are fixed by the payouts and the probabilities. Nothing about the last roll changes the probability of the next one.
| Game | Does card removal change future odds? | Can the player act on the change? | Why counting does or does not apply |
|---|---|---|---|
| Blackjack | Yes, until the shuffle | Yes, by changing bet size and playing decisions | Player decisions interact with a changing deck composition |
| Baccarat | Yes, cards are removed | Almost not at all | Fixed drawing rules leave no meaningful decision to adjust |
| Roulette | No | No | Independent spins from a fixed wheel |
| Craps | No | No | Independent rolls from fixed dice |
Baccarat removes cards but still resists counting
Baccarat looks like it should be countable because it is a comparing card game between the player and banker hands where totals are taken modulo 10, and cards are dealt from a shoe. The problem is not that the shoe has no memory. The problem is that the player has almost no decision to make after the cards are seen. The player chooses whether to bet on the player hand, the banker hand or a tie, and then the drawing rules play out automatically. A counter can sometimes detect that the remaining shoe slightly favours one side or the other, but the effect is small and the opportunity is rare. The banker, player and tie bets carry different house edges, with the tie bet the highest, and those edges stay close to their long-run values even when the deck composition shifts a little.
In blackjack, the player's choices multiply the value of the information. The player decides whether to hit, stand, double or split, and those choices are worth more when the remaining cards are known to be rich in tens and aces. Baccarat gives the player no comparable lever. The information exists but cannot be converted into a meaningful change in expected value.
Where the edge comes from and what counting changes
House edge and RTP are complements: they sum to 100% for the same game under the same rule set. A blackjack game with a published basic strategy has an expected loss per hand that follows from its specific rule set, including the deck count, the dealer stand rules and the doubling and splitting rules. A counter is not removing that edge by magic. The counter is betting more when the remaining shoe has shifted the edge in the player's direction and less when it has not. Over many hands, the average bet is placed at a slightly better moment than a flat bettor's would be.
That shift is narrow. Suppose, as an example, a game returns 99.5% to a basic-strategy player under one rule set, leaving a house edge of 0.5%. A counter might move the edge by a fraction of a percentage point when the count is strongly positive, and only on the hands where the count is positive enough to justify a larger bet. The counter still loses many individual hands. The edge is a long-run average over a very large number of hands, not a per-session guarantee.
Short sessions look nothing like the long-run average
A player who counts cards for one evening is seeing variance, not expectation. The expected value of a hand is the average result if that hand were played an enormous number of times from the same deck state. A single session contains too few hands for that average to dominate. The player can play correctly, raise bets at the right moments and still finish down, because the cards can fall against a favourable deck for hours.
This is the same reason a non-counter can win on any given night. Short-run outcomes and long-run averages diverge, and the shorter the session, the larger the divergence can be. Counting does not shorten that timeline. It only shifts the average slightly in the player's favour, and only when the player executes it accurately across enough hands for the shift to accumulate.
Decisions per hour change total exposure
Blackjack hands play quickly. The number of decisions per hour multiplies the total amount wagered, and total wagered is what the house edge acts on. A player betting a fixed amount per hand at a table dealing many hands per hour puts far more money through the game than the same player would at a slower table. The expected loss is the product of the total amount wagered and the house edge, so the speed of the game changes the size of the expected loss even when the edge itself stays the same.
A counter raising bets in a positive count is also raising the amount wagered per hour, which means the counter's bankroll swings harder. The edge may be slightly better on those raised bets, but the variance is also larger. A short run of bad luck on large bets can erase a long run of small gains, which is why counting requires a bankroll large enough to absorb the swings and a time horizon long enough for the average to matter.
Why progression systems do not replace counting
Betting progressions, such as doubling after a loss, change the pattern of wins and losses without changing the expected value of the underlying game. A progression redistributes outcomes: many small wins and a few large losses, or the reverse, depending on the pattern. It does not remove the house edge on any individual bet, and it does not change the probability of the next hand. The next hand in blackjack is not independent in the same way as a roulette spin, because the deck composition changes, but the progression itself is not reading that composition. It is only reacting to whether the last hand won or lost.
Counting is different because it reacts to the cards that have been removed, not to the player's recent results. That is the line that separates information from superstition. A counter's larger bet is justified by a changed deck state. A progression player's larger bet is justified by nothing except the previous outcome, and the previous outcome does not change the edge on the next hand.
What the player can and cannot control
The player controls bet selection, the decision to follow a published basic-strategy chart for the specific rule set in use, the speed of play, the session length and the bankroll size. Each of those changes the experience of the game and the total exposure. None of them changes the built-in edge on a negative-expectation game into a guaranteed win. Counting, when executed accurately, can shift the average slightly in the player's favour under favourable deck conditions, but it does not make any single hand predictable and it does not remove variance.
No betting pattern changes the built-in edge of a negative-expectation game. Progression systems redistribute outcomes rather than remove the edge. Counting is not a progression system, but it is also not a shortcut. It is a narrow, demanding response to a specific game structure, and that structure is the reason the question is asked about blackjack and not about the wheel or the dice.
FAQ
Can card counting work in blackjack variants with continuous shufflers?
Continuous shufflers return discards to the shoe before a meaningful number of cards has been removed, which prevents the deck composition from shifting far enough to exploit. The information a counter needs never accumulates, so the method has no practical effect under that equipment.
Is counting cards the same as cheating?
Counting uses information that is visible to every player at the table. It is not device use, collusion or tampering with the game. Venues may still decline to serve a player they believe is counting, and they may change the rules or shuffle more often, but the act itself is a mental strategy rather than a manipulation of the equipment.
Why do casinos care about counting if the edge is so small?
A small edge applied to a large amount of wagered money becomes a meaningful expected profit over time. The venue's concern is not any single hand but the long-run average across many sessions, and a skilled counter who bets more in favourable counts is placing a larger share of total wagers at a slightly better price.
Gamble responsibly. If you or someone you know has a gambling problem, call the National Problem Gambling Helpline at the helpline printed in the footer below or visit NCPGambling.org.
Play responsibly
Gambling should be treated as paid entertainment, never as a way to earn income or recover losses. 1-800-MY-RESET (1-800-697-3738)
18+ or 21+ depending on where you are; follow the minimum age that applies to you.
This article is general information about how these mechanics work. It is not legal advice and not a recommendation to gamble or to use any particular operator. Availability and legality differ by jurisdiction — check the rules that apply where you are.