How to count cards in blackjack (and what it’s honestly worth)
Card counting is the only legitimate way to play blackjack with an edge over the casino. It is also legal, widely misunderstood, smaller in profit than the movies suggest, and almost useless online. This guide covers the real thing: the Hi-Lo system, true-count conversion, bet spreads, the strategy deviations that matter — and the honest arithmetic of what a counter earns.
Why counting works at all
Basic strategy assumes a fresh shoe. But a real shoe has memory: when the cards already dealt were mostly small, the remaining shoe is rich in tens and aces — which helps the player (more blackjacks at 3:2, dealer busts stiff hands more often, doubles connect more). When the remaining shoe is poor in tens, the edge swings the other way. Counting is nothing more than tracking that imbalance with a running number and betting big only when the shoe favors you.
The Hi-Lo system
Hi-Lo assigns a value to every card you see — yours, other players’, the dealer’s:
| Cards | Count value |
|---|---|
| 2, 3, 4, 5, 6 | +1 |
| 7, 8, 9 | 0 |
| 10, J, Q, K, A | −1 |
Start at 0 after the shuffle and keep a running count. A positive count means small cards have left the shoe and the remainder is ten-rich. To compare counts across the shoe, convert to a true count: divide the running count by the number of decks you estimate remain. Running +6 with three decks left is a true count of +2. The true count is the number every decision hangs on: as a rule of thumb, each point of true count shifts the player’s expectation by roughly half a percent, so at about TC +1 a good game reaches break-even, and beyond it the player is printing a small, real edge.
Betting the count
The edge only pays if your money follows it: bet the table minimum when the count is neutral or negative, and scale up as the true count climbs — a 1-to-8 or 1-to-12 spread is typical. This is also exactly what surveillance looks for, which is why practical counting is as much about camouflage and game selection as arithmetic. Penetration matters more than almost anything: a dealer who cuts off two of eight decks kills more EV than a sloppy count ever will.
Deviations: when the count overrules basic strategy
At certain counts the mathematically correct play flips. These index plays — the famous "Illustrious 18" — are worth far more than perfect play in obscure cells. The first three cover most of the value:
| Situation | Deviation |
|---|---|
| Insurance offered | Take it at TC +3 or higher (the only time insurance is ever correct) |
| 16 vs 10 | Stand at TC 0 or higher, hit below |
| 15 vs 10 | Stand at TC +4 or higher |
| 12 vs 3 | Stand at TC +2 |
| 12 vs 2 | Stand at TC +3 |
| 10 vs 10 | Double at TC +4 |
| 10-10 vs 6 | Split at TC +4 (yes, really — but only then) |
| 11 vs Ace | Double at TC +1 |
Each linked page shows the basic-strategy math for the neutral shoe, so you can see exactly how thin the margin is that the count tips over.
The honest arithmetic
A skilled Hi-Lo counter with a good game (S17, DAS, deep penetration) and a sensible spread earns a long-run edge of roughly 0.5–1% of total money wagered. That is a real edge — casinos ban people over it — but it arrives wrapped in brutal variance: hundred-unit downswings are routine, and the hourly expectation for a $10–$25 bettor is closer to a modest wage than a heist. Counting is legal everywhere (it is thinking, not a device), but private casinos may refuse your action; using any counting device or app at a live table, however, is a crime in most jurisdictions.
Why counting dies online
Online RNG blackjack shuffles every hand — the shoe has no memory, so there is nothing to count. Live-dealer studios deal from real shoes but typically cut them shallow and shuffle early, and continuous shuffling machines (CSMs) at many tables return cards to the shoe constantly. A few deeply-dealt live tables offer marginal opportunities, but nothing like a good land-based game. If a bot or seller promises you "online card counting signals", you now know exactly what that is worth.
How to learn it
Three stages: first, drill basic strategy to 100% — deviations are worthless on a broken foundation. Second, practice the running count with a deck of cards until you can count one down in under 30 seconds (the classic benchmark). Third, add true-count conversion and the first few deviations. Our solver-backed trainer drills the strategy layer today; a dedicated counting drill is on the roadmap.