How card counting reads a shoe
Every card dealt changes what is left in the shoe. Take out a five and the rest of the shoe holds slightly more tens and aces in proportion. High cards favour the player on average, because a natural pays 3:2 and the dealer must draw to stiff totals. Low cards favour the dealer, who completes more hands without busting.
A count does not predict the next card. The next card is still drawn at random from whatever remains. What the count measures is composition, the share of high cards left compared with a fresh shoe. That share moves a few percentage points at most, and it moves only when many cards have gone and few remain.
The Hi-Lo tags, rank by rank
The Hi-Lo system gives each rank one of three tags. Wizard of Odds and the Wikipedia article on card counting give the same values: the 2 to 6 are +1, the 7 to 9 are 0, and the ten-value cards and aces are -1. The table below adds how many of each group a six-deck shoe holds.
| Rank | Hi-Lo tag | Cards in a six-deck shoe | Effect on the cards left when one is dealt |
|---|---|---|---|
| 2, 3, 4, 5, 6 | +1 | 120 | Shoe gets richer in high cards |
| 7, 8, 9 | 0 | 72 | No change counted |
| 10, J, Q, K | -1 | 96 | Shoe gets poorer in high cards |
| Ace | -1 | 24 | Shoe gets poorer in high cards |
The system is balanced. A full shoe has 120 cards tagged +1 and 120 tagged -1, so a count run through every card ends at zero. That is a useful self-check when practising with a real deck at home. The 72 neutral cards are ignored entirely, which keeps the arithmetic to a single addition per card.
Running count and true count on a real shoe
Take a six-deck shoe at a land casino table dealt to a cut card, with no continuous shuffler. Two decks, 104 cards, have been dealt. Of those, 48 were low cards from 2 to 6, 22 were sevens, eights or nines, and 34 were tens or aces. The running count is 48 minus 34, which gives +14.
| Group | At the shuffle | Dealt so far | Still in the shoe |
|---|---|---|---|
| 2 to 6 (+1) | 120 | 48 | 72 |
| 7 to 9 (0) | 72 | 22 | 50 |
| Tens and aces (-1) | 120 | 34 | 86 |
| Total | 312 | 104 | 208 |
The true count divides the running count by the decks not yet dealt. Here 208 cards remain, four decks, so +14 becomes a true count of +3.5. The same +14 with one deck left would be +14, and with five decks left it would be +2.8. A running count on its own says almost nothing about the shoe.
What actually changed is visible in the last column. At the shuffle, tens and aces made up 120 of 312 cards, or 38.5% of the shoe. Now they are 86 of 208, or 41.3%. That shift of under three points is what a true count of +3.5 describes, and it took two full decks of play to produce.
The count also changes a few decisions. Wizard of Odds publishes the Illustrious 18 index plays for the Hi-Lo: insurance becomes the correct choice at a true count of +3 or higher, and hard 16 against a ten is stood at 0 or above. Hard 15 against a ten waits until +4, so at +3.5 it is still a hit.
What the published Hi-Lo efficiency figures mean
Peter Griffin’s The Theory of Blackjack (1978) split a count’s value into two measures, and figures published by QFIT give the Hi-Lo a betting correlation of .97, a playing efficiency of .51 and an insurance correlation of .76. The betting figure says how closely the count tracks the shoe’s real value for sizing a bet.
A playing efficiency of .51 means the count captures roughly half of what perfect knowledge of the remaining cards would gain on strategy decisions. The insurance correlation of .76 says the count is a fairly good guide to the one bet that depends only on tens. None of the three says anything about the house edge on its own.
The simulation figures that follow come from Wizard of Odds, and they rest on stated conditions: six decks, dealer stands on soft 17, surrender and double after split allowed, resplits to four hands including aces, one betting spot, bets raised only after a win and lowered only after a loss. Few tables meet that combination.
| Bet spread | Decks dealt of 6 | Index plays | Simulated result per initial bet | Std. dev. (units) | Average bet (units) |
|---|---|---|---|---|---|
| 1 to 5 | 4 | Illustrious 18 + Fab 4 | 0.157% | 1.50 | 1.35 |
| 1 to 15 | 4 | Illustrious 18 + Fab 4 | 0.578% | 2.67 | 1.73 |
| 1 to 5 | 5 | Illustrious 18 + Fab 4 | 0.469% | 1.70 | 1.47 |
| 1 to 15 | 5 | Illustrious 18 + Fab 4 | 1.147% | 3.49 | 2.10 |
| 1 to 15 | 5 | Full index table | 1.182% | 3.54 | 2.11 |
The pattern matters more than any row. Moving the cut from four decks dealt to five doubles the figure at the same spread, while adding every index number on top of the Illustrious 18 changes it by a few hundredths. Penetration and bet spread do the work. The strategy deviations are a refinement, worth little without the other two.
The swings dwarf the average. In the 1-to-15 row with five decks dealt, the standard deviation is 3.49 units per hand, against an average result of about 0.024 units. By our arithmetic on those figures, around 21,000 hands pass before the average equals one standard deviation. Over any evening, the count’s effect is invisible in the results.
A counter still plays at a house edge most of the time. At neutral or negative counts the shoe is ordinary or worse, and on a standard six-deck shoe that is the 0.41% our maths engine computes for its rules. The simulated average comes from larger bets at the minority of high counts outweighing the minimum bets everywhere else.
The bet spread is also the visible part. Wizard of Odds calls a 1-to-15 spread about the most aggressive a six-deck player should use, and its simulation only raised bets after wins to avoid attention. Smaller spreads shrink the figures sharply, as the first row shows. The arithmetic of the count is simple; the conditions it needs are not.
The published Hi-Lo figures assume the player already makes the basic strategy play on every hand, because a count only adjusts a handful of decisions at the margin. The blackjack trainer drills those base decisions on six-deck rules with feedback after each hand. It uses no money and keeps no score between visits.
Card counting in online blackjack: why RNG and continuous shuffles stop it
Computer-dealt blackjack in our Ontario records reshuffles between hands. Specialist reviews of Games Global’s Atlantic City state that its eight decks are shuffled before each game, and reviews of Vegas Single Deck that the cards are shuffled before every hand. Every deal starts from a complete shoe, so the count resets before it can move.
That makes basic strategy the exact best play on those tables, not an approximation. A chart is calculated from a freshly shuffled shoe, which is precisely what each RNG hand is dealt from. The strategy chart for the matching rule set and the house edge calculator describe those games completely, and nothing a player tracks between hands changes either figure.
One computer table in our records works differently. Pragmatic Play describes Blackjack X as dealing eight decks with no reshuffle until the cut card, with the cut and the shuffle animated on screen. The cut card appears after about four decks, so a count can build, but eight decks move the true count slowly.
Continuous shuffling machines
A continuous shuffler takes discards back after each round and mixes them into the cards still to be dealt. The shoe never runs down, so its composition stays close to a fresh one. Wizard of Odds put it plainly: “If you’re any kind of counter at all I would forget about playing against a CSM, it isn’t worth the bother.”
Evolution’s Always 6 Blackjack, dealt at ToonieBet, runs its six-deck shoe through a continuous shuffler, so a count there has nothing to build on. The other live tables in the register deal from shoes replaced at a cut card. Basic strategy is unchanged by a shuffler, because it already assumes a freshly shuffled shoe.
Live dealer shoes and the shuffle point
Live blackjack is dealt from a physical shoe, so the cards can be tracked. Whether that has any value depends on penetration, how much of the shoe is dealt before the shuffle. Wizard of Odds lists Evolution’s standard live blackjack at 50%, and Blackjack Review describes eight-deck live shoes replaced once four decks are dealt.
Infinite Blackjack is the clearest case. Evolution deals it from eight decks, every player shares the same first two cards, and the number of seats is unlimited, so a count describes one shoe for everyone. Its published return of 99.47% gives a 0.53% house edge, and the Infinite Blackjack page covers the rest of its rules.
The one Ontario procedure in print points the same way. The AGCO’s rules of play for fair and exhibition event licences, which cover charitable events rather than casinos, place the cut card at approximately halfway and require a reshuffle once the cutting card is reached and the last hand is completed.
The arithmetic shows why depth matters on eight decks. If half an eight-deck shoe were dealt, four decks would remain at the very last hand, and a running count of +12 would be needed for a true count of +3. Earlier in the shoe the same count means less. At the live tables basic strategy remains the standard.
What changes the count’s value
The same count is worth very different amounts depending on the table. The factors below come from the simulation figures above, the shuffle methods stated in each rules file and the penetration figures reference sites give for live shoes. Each row shows the effect on the count and the Ontario tables where it applies.
| Factor | What it does to the count | What we found for Ontario tables |
|---|---|---|
| Number of decks | More decks divide the running count further, so the true count moves more slowly | Eight decks on Infinite Blackjack, ONE Blackjack and Blackjack X; six on Always 6 |
| Penetration | Deeper cuts leave fewer decks at the end, where true counts swing most | About 50% on Evolution’s standard live shoes; AGCO fair rules cut at about halfway |
| Reshuffle every hand | Count resets to zero before each deal | Games Global’s Vegas Single Deck and Atlantic City |
| Continuous shuffler | Discards return, so the shoe never runs down | Evolution’s Always 6 Blackjack |
| Shoe dealt to a cut card | Count can build, value depends on the cut | Blackjack X after about four of eight decks; live shoes around halfway |
| Common shared hand | Same cards for every seat; no effect on the maths of the shoe | Infinite Blackjack, eight decks |
| Base house edge | The starting point a count has to move | 0.31% Vegas Single Deck; 0.53% Infinite Blackjack |
Read down the last column and the pattern is plain. The computer tables reset every hand, Always 6 uses a machine, and the shoe tables cut around the middle of an eight-deck shoe. That leaves the base house edge, which every table carries, as the number a player can actually compare.
Counting and the Criminal Code of Canada
Section 209 of the Criminal Code reads: “Every person who, with intent to defraud any person, cheats while playing a game or in holding the stakes for a game or in betting is guilty of (a) an indictable offence and liable to imprisonment for a term of not more than two years; or (b) an offence punishable on summary conviction.”
The section turns on cheating with intent to defraud, and it does not mention counting. Keeping track of cards dealt face up uses only information the table shows every player, so on this text card counting is legal in Canada. In 1995 a Windsor court acquitted three players charged with cheating over team play.
The law on premises is a separate matter. A casino controls who it deals to and can ask a person to leave, and in Ontario the Trespass to Property Act is the statute venues rely on. Refusing service is a property right, not a finding of cheating. This is a plain reading of the texts, not legal advice.
Online in Ontario, the question rarely arises, because the tables that state a method reshuffle each hand. At a shoe table, the check worth making is in the game rules screen: the deck count and where the cut card sits. Then play the chart for the published rules, which is the realistic standard at every table.