Blackjack Card Counting: How Hi-Lo Works and What It Really Earns

Card counting is the one well-known way to get a real mathematical edge over a casino game. It is also slower, harder and less profitable than most people think. This guide explains the Hi-Lo count step by step, from the card values to the true count and the bet spread, and then shows what it is actually worth. For that we simulated 3.3 billion rounds of six-deck blackjack under the same rules as our free blackjack game.

The short version: with good rules and a deep shoe, a counter who makes no mistakes wins about 1 betting unit per 100 hands. You can practise with the count trainer further down this page, no real money involved.

0.36%
House edge in our game with basic strategy
+0.6
Points of player edge per step of the true count
≈ $8
Expected win per hour with a $10 to $80 bet spread
35%
Of perfect counters still behind after 125 hours

1. Why card counting works

In roulette every spin starts from zero: the wheel has no memory. Blackjack is different, because the cards come out of a shoe and the cards that have been played are gone until the next shuffle. That means the composition of the remaining cards changes from round to round, and with it the odds.

Tens and Aces left in the shoe are good for you, and low cards are good for the dealer. There are four reasons:

  • More blackjacks. A blackjack needs a ten and an Ace. You and the dealer get them equally often, but yours pays 3:2 and the dealer’s only takes your bet.
  • The dealer busts more. The dealer must hit 12 to 16. When the shoe is full of tens, those hits bust more often. You can stand on the same hands; the dealer cannot.
  • Doubling pays better. When you double on 10 or 11, a ten for your one card is what you want.
  • Insurance becomes a good bet. Normally insurance is a losing side bet. If more than a third of the remaining cards are tens, it turns profitable.

Low cards work the other way: they turn the dealer’s stiff hands into 17 to 21. So a counter does not need to remember individual cards. It is enough to keep track of the balance between low and high cards that have already been played.

2. Learn basic strategy first

Counting is built on top of basic strategy, the mathematically best way to play every hand. With basic strategy, the house edge in our game is 0.36%. Players who decide on gut feeling often give away 1% to 2% or more, which is more than counting can ever win back.

So the order is fixed: first basic strategy until every decision is automatic, then the count. Our blackjack guide explains the rules and the strategy, and the game on that page has a coach that shows the basic strategy move for every hand.

3. The Hi-Lo count step by step

Hi-Lo is the system most counters use. It was published by Harvey Dubner in 1963, shortly after Edward Thorp’s Beat the Dealer, and it is still the standard because it is simple and still captures most of what a more complex count would.

Step 1: the card values

2♣3♥4♠5♦6♣Low cards 2 to 6: count +1
7♥8♠9♦Middle cards 7 to 9: count 0
10♠J♥Q♣K♦A♠Tens, faces and Aces: count −1

The values add up to zero over a full deck: five cards count +1, five count −1 and three count 0. That is why a freshly shuffled shoe always starts at zero.

Step 2: the running count

Start at zero after the shuffle and add the value of every card you see: your cards, the other players’ cards, the dealer’s upcard, the hole card when it is turned over and every card drawn. Here is one round:

CardValueCount
K♠−1−1
5♥+10
6♦+1+1
2♣+1+2
9♠0+2
A♥−1+1
3♥+1+2
4♠+1+3
Q♦−1+2

After this round the running count is +2. Experienced counters do not add card by card. They cancel pairs: a King and a 5 together make 0, so they skip both. That makes it fast enough to keep up with a real dealer.

Step 3: the true count

A running count of +6 means much more when only one deck is left than when five decks are left. The true count corrects for that: it is the running count per deck that has not been played yet.

True Count

True count = running count ÷ decks remaining

Example: running count +9, about 3 decks left in the shoe → true count +3.

Decks remaining: estimate them from the discard tray. In a 6-deck shoe with 2 decks in the tray, 4 decks are left.

Rounding: most counters round down. In this guide, “true count +2” means anything from +2 up to just under +3.

Step 4: bet more when the count is high

The true count tells you when you have the edge. You make money only by betting more at those moments and as little as possible the rest of the time. The ratio between your smallest and largest bet is called the bet spread. These are the two spreads we simulated:

True countSpread 1–8Spread 1–12With $10 units (1–8)
+1 or lower1 unit1 unit$10
+22 units2 units$20
+34 units4 units$40
+46 units8 units$60
+5 or higher8 units12 units$80
Bet ramps used in our simulation. True counts are rounded down.

4. What our simulation shows

We wrote our own blackjack simulator and played 3.3 billion rounds with the rules of our game: 6 decks, dealer stands on soft 17, double on any two cards and after splits, split up to four hands, late surrender, blackjack pays 3:2 and a cut card after 75% of the shoe. The simulated player uses perfect basic strategy, counts Hi-Lo without mistakes and uses the playing deviations from section 5. How we did it is described at the end of this page.

Your edge at each true count

Player edge by true count
Player edge with Hi-Lo + deviationsHouse edgeBasic strategy only
Player edge by Hi-Lo true count, 6 decks −2% −1% 0% +1% +2% +3% +4% +5% True count −4 (from -4 up to just under -3): player edge −2.09% with Hi-Lo and deviations, −2.34% with basic strategy only · 3.9% of all rounds −4 3.9 True count −3 (from -3 up to just under -2): player edge −1.59% with Hi-Lo and deviations, −1.68% with basic strategy only · 6.7% of all rounds −3 6.7 True count −2 (from -2 up to just under -1): player edge −1.05% with Hi-Lo and deviations, −1.14% with basic strategy only · 12.4% of all rounds −2 12.4 True count −1 (from -1 up to just under 0): player edge −0.57% with Hi-Lo and deviations, −0.55% with basic strategy only · 18.9% of all rounds −0.6% −1 18.9 True count 0 (from 0 up to just under 1): player edge −0.09% with Hi-Lo and deviations, −0.10% with basic strategy only · 26.7% of all rounds −0.1% 0 26.7 True count +1 (from 1 up to just under 2): player edge +0.52% with Hi-Lo and deviations, +0.51% with basic strategy only · 11.8% of all rounds +1 11.8 True count +2 (from 2 up to just under 3): player edge +1.13% with Hi-Lo and deviations, +1.05% with basic strategy only · 6.5% of all rounds +1.1% +2 6.5 True count +3 (from 3 up to just under 4): player edge +1.73% with Hi-Lo and deviations, +1.61% with basic strategy only · 3.7% of all rounds +3 3.7 True count +4 (from 4 up to just under 5): player edge +2.43% with Hi-Lo and deviations, +2.06% with basic strategy only · 2.1% of all rounds +2.4% +4 2.1 True count +5 (from 5 up to just under 6): player edge +3.13% with Hi-Lo and deviations, +2.46% with basic strategy only · 1.2% of all rounds +5 1.2 True count +6 (from 6 up to just under 7): player edge +3.76% with Hi-Lo and deviations, +2.89% with basic strategy only · 0.6% of all rounds +3.8% +6 0.6 True count +7 (from 7 up to just under 8): player edge +4.55% with Hi-Lo and deviations, +3.23% with basic strategy only · 0.4% of all rounds +7 0.4 True count +8 (from 8 up to just under 9): player edge +5.13% with Hi-Lo and deviations, +3.66% with basic strategy only · 0.2% of all rounds +5.1% +8 0.2 TC % of rounds
Expected win or loss per hand at each true count (rounded down), 6 decks, our game’s rules. Grey dots: the same player without playing deviations. Bottom row: how often each true count occurs. Hover over a bar for details. Source: GamblingSociety simulation, 300 to 400 million rounds per strategy.

Each step of the true count is worth about 0.6 percentage points. At a true count of 0 the game is almost even (−0.1%), from +1 you are ahead, and at +2 you have about +1.1%. The catch is in the bottom row: a true count of +2 or higher comes up in only about 15% of rounds. Most of the time you play with the house ahead and bet the minimum.

What it earns per hour

How you playUnits won per 100 handsPer hour with $10 unitsBankroll for a 5% risk of ruinUntil skill outweighs luck
Basic strategy, flat bet−0.36−$2.87No bankroll is enoughNever
Hi-Lo with deviations, flat bet−0.26−$2.09No bankroll is enoughNever
Hi-Lo with deviations, spread 1–8+0.96+$7.71834 units ($8,300)≈ 722 hours
Hi-Lo with deviations, spread 1–12+1.47+$11.74903 units ($9,000)≈ 514 hours
Assumes 80 hands per hour. “Until skill outweighs luck” is the point where the expected win equals one standard deviation of the result (N0). Source: GamblingSociety simulation.

Two things stand out. First, counting without changing your bet is almost worthless: the playing deviations alone only cut the house edge from 0.36% to 0.26%. The money comes from the bet spread. Second, even a perfect counter with a $10 to $80 spread earns about $8 an hour on average, needs a bankroll of around $8,000 and plays over 700 hours before the result reflects skill more than luck.

How long until counting shows

Averages hide how long the swings last. So we also simulated 2,000 complete “careers” for each strategy: every simulated player plays 200,000 hands in a row, and we checked how many were behind at different points along the way.

Share of players who are behind after a number of hands
Basic strategy, flat betHi-Lo, bet spread 1–8Hi-Lo, bet spread 1–12
Share of simulated players who are behind after a given number of hands 0% 25% 50% 75% 100% 1k ≈12 h 5k ≈62 h 10k ≈125 h 20k ≈250 h 50k ≈625 h 100k ≈1,250 h 200k ≈2,500 h Hands played (hours at 80 hands per hour) Basic strategy, flat bet, after 1,000 hands: 51.2% behind · median result −2 units (90% of players between −63 and +56) Basic strategy, flat bet, after 5,000 hands: 59.0% behind · median result −16 units (90% of players between −149 and +115) Basic strategy, flat bet, after 10,000 hands: 62.5% behind · median result −36 units (90% of players between −217 and +151) Basic strategy, flat bet, after 20,000 hands: 67.0% behind · median result −68 units (90% of players between −331 and +194) Basic strategy, flat bet, after 50,000 hands: 76.8% behind · median result −186 units (90% of players between −609 and +244) Basic strategy, flat bet, after 100,000 hands: 83.3% behind · median result −362 units (90% of players between −957 and +215) Basic strategy, flat bet, after 200,000 hands: 91.6% behind · median result −709 units (90% of players between −1,547 and +137) 92% Hi-Lo, bet spread 1–8, after 1,000 hands: 45.0% behind · median result +10 units (90% of players between −112 and +128) Hi-Lo, bet spread 1–8, after 5,000 hands: 39.3% behind · median result +43 units (90% of players between −212 and +309) Hi-Lo, bet spread 1–8, after 10,000 hands: 34.7% behind · median result +96 units (90% of players between −280 and +457) Hi-Lo, bet spread 1–8, after 20,000 hands: 26.6% behind · median result +194 units (90% of players between −380 and +723) Hi-Lo, bet spread 1–8, after 50,000 hands: 18.6% behind · median result +468 units (90% of players between −371 and +1,326) Hi-Lo, bet spread 1–8, after 100,000 hands: 9.4% behind · median result +975 units (90% of players between −229 and +2,144) Hi-Lo, bet spread 1–8, after 200,000 hands: 2.5% behind · median result +1,921 units (90% of players between +302 and +3,669) 2.5% Hi-Lo, bet spread 1–12, after 1,000 hands: 43.6% behind · median result +14 units (90% of players between −146 and +173) Hi-Lo, bet spread 1–12, after 5,000 hands: 35.6% behind · median result +72 units (90% of players between −264 and +428) Hi-Lo, bet spread 1–12, after 10,000 hands: 30.1% behind · median result +144 units (90% of players between −335 and +638) Hi-Lo, bet spread 1–12, after 20,000 hands: 22.2% behind · median result +300 units (90% of players between −337 and +956) Hi-Lo, bet spread 1–12, after 50,000 hands: 12.7% behind · median result +759 units (90% of players between −303 and +1,811) Hi-Lo, bet spread 1–12, after 100,000 hands: 5.7% behind · median result +1,495 units (90% of players between −72 and +2,944) Hi-Lo, bet spread 1–12, after 200,000 hands: 1.2% behind · median result +2,991 units (90% of players between +744 and +5,113) 1.2%
Share of 2,000 simulated players per strategy whose total result is below zero. Perfect play assumed. Hover over a point for the typical range of results. Source: GamblingSociety simulation.

After 10,000 hands, about 125 hours at the table, 35% of perfect counters with a 1–8 spread are still behind. After 50,000 hands it is 19%, and only after about 200,000 hands are almost all of them ahead. For the basic strategy player it goes the other way: after 200,000 hands, 92% are behind. That is the house edge at work.

Game conditions matter more than the system

Expected win per 100 hands under different table conditions
Bet spread 1–8Bet spread 1–12
Expected win per 100 hands in betting units, by game conditions −1.5 −1.0 −0.5 0.0 +0.5 +1.0 +1.5 +2.0 +2.5 6 decks, 50% dealt (typical live dealer) 6 decks, 50% dealt (typical live dealer), spread 1–8: +0.07 units per 100 hands (edge on money bet +0.06%) +0.07 6 decks, 50% dealt (typical live dealer), spread 1–12: +0.16 units per 100 hands (edge on money bet +0.13%) +0.16 6 decks, 65% dealt 6 decks, 65% dealt, spread 1–8: +0.50 units per 100 hands (edge on money bet +0.37%) +0.50 6 decks, 65% dealt, spread 1–12: +0.79 units per 100 hands (edge on money bet +0.55%) +0.79 6 decks, 75% dealt (our game) 6 decks, 75% dealt (our game), spread 1–8: +0.96 units per 100 hands (edge on money bet +0.66%) +0.96 6 decks, 75% dealt (our game), spread 1–12: +1.47 units per 100 hands (edge on money bet +0.91%) +1.47 6 decks, 83% dealt 6 decks, 83% dealt, spread 1–8: +1.47 units per 100 hands (edge on money bet +0.94%) +1.47 6 decks, 83% dealt, spread 1–12: +2.23 units per 100 hands (edge on money bet +1.26%) +2.23 Dealer hits soft 17, 75% Dealer hits soft 17, 75%, spread 1–8: +0.69 units per 100 hands (edge on money bet +0.47%) +0.69 Dealer hits soft 17, 75%, spread 1–12: +1.18 units per 100 hands (edge on money bet +0.73%) +1.18 Blackjack pays 6:5, 75% Blackjack pays 6:5, 75%, spread 1–8: −1.16 units per 100 hands (edge on money bet −0.79%) −1.16 Blackjack pays 6:5, 75%, spread 1–12: −0.91 units per 100 hands (edge on money bet −0.56%) −0.91
Expected win per 100 hands in betting units, Hi-Lo with deviations, all other rules as in our game. “Dealt” is how much of the shoe is played before the shuffle (penetration). Source: GamblingSociety simulation.

The same counter with the same spread wins almost three times as much in a shoe dealt to 83% than in one dealt to 65%. At 50%, which is common on live-dealer tables, the edge is almost gone (+0.16 units per 100 hands with a 1–12 spread). And at a table where blackjack pays 6:5, counting cannot win at all: even with a 1–12 spread the player loses 0.91 units per 100 hands. A dealer who hits soft 17 costs a fifth to a quarter of the win rate.

5. Playing deviations: the Illustrious 18 and the Fab 4

Basic strategy is right for an average shoe. When the true count is high or low, some decisions change. The most valuable of these changes were collected by Don Schlesinger as the “Illustrious 18” and, for surrender, the “Fab 4”. Insurance at a true count of +3 or higher is the single most valuable one.

Your handDealer showsDeviation
InsuranceAceTake it at +3 or higher
1610Stand at 0 or higher (when you cannot surrender)
1510Stand at +4 or higher (when you cannot surrender)
10, 105Split at +5 or higher
10, 106Split at +4 or higher
1010Double at +4 or higher
10AceDouble at +4 or higher
123Stand at +2 or higher
122Stand at +3 or higher
11AceDouble at +1 or higher
92Double at +1 or higher
97Double at +3 or higher
169Stand at +5 or higher (when you cannot surrender)
132Hit below −1
124Hit below 0
125Hit below −2
126Hit below −1
133Hit below −2
1410Surrender at +3 or higher
1510Surrender at 0 or higher, otherwise hit
159Surrender at +2 or higher
15AceSurrender at +1 or higher
Hi-Lo index plays for 6 decks, dealer stands on soft 17, as used in our simulation. The last four rows are the Fab 4 surrender plays and apply only to your first two cards.

In our simulation, these 22 plays raised the win rate with a 1–8 spread from +0.62 to +0.96 units per 100 hands, a gain of about half. Learn them after the count is automatic, starting with insurance and 16 against 10.

6. Bankroll and risk of ruin

An edge of about 1% does not protect you from long losing streaks. Your bankroll has to be big enough to survive them, or you go broke before the edge can show. Risk of ruin is the chance of losing the whole bankroll. This is what we found for a counter with a 1–8 spread over 200,000 hands (about 2,500 hours):

BankrollWith $10 unitsRuined within 200,000 hands
200 units$2,00047.0%
400 units$4,00022.4%
800 units$8,0005.4%
1,200 units$12,0000.9%
Share of 2,000 simulated counters (Hi-Lo with deviations, spread 1–8) whose losses reached the full bankroll at some point. Perfect play assumed. Source: GamblingSociety simulation.

Even 400 units, $4,000 with a $10 minimum, leaves more than a one-in-five chance of going broke. Professional counters therefore often keep a bankroll of around 1,000 units or more, and they lower their bets when the bankroll shrinks. If you want to size bets from your edge and bankroll in a systematic way, our Kelly criterion calculator shows the principle.

7. How to practise

Counting is a skill like typing: you get faster by repetition, not by understanding. These drills are the usual way to build it up:

  • Count down a deck. Shuffle one deck, hold back one card, then turn the rest over one by one and count. The result must match the value of the card you held back (with the sign reversed). Aim for under 25 seconds.
  • Count in pairs. Turn over two cards at a time and learn to see pairs that cancel out (a 4 and a Queen) as zero.
  • Convert to a true count. Put a few decks of discards in a pile, guess the decks remaining, and divide. Do it until it takes a second.
  • Play with distractions. Talk, listen to music or watch TV while you count. In a casino nobody waits for you.
  • Play real rounds. Use the trainer in our game below: it deals from a 6-deck shoe, asks for the running count at random moments and checks your answer.

8. Card counting in a real casino

The math above assumes that the casino lets you play. In practice, casinos watch for counters, and the main giveaway is the bet spread: a player who bets $10 for twenty minutes and then suddenly $80 is easy to spot. These are the usual countermeasures:

  • Shuffling early. The dealer shuffles as soon as you raise your bet, or the cut card is placed further forward. As the chart above shows, this alone cuts the win rate sharply.
  • Flat betting. You are told that you can keep playing, but only with the same bet every hand. Without the spread, counting does not pay.
  • Barring. You are asked to stop playing blackjack or to leave the casino.
  • Shuffling machines. Continuous shuffling machines put the used cards straight back into the shoe. The count never builds up, so counting is useless at these tables.
  • No mid-shoe entry. Back-counting (watching a table and joining only when the count is high, also called “wonging” after Stanford Wong) is very effective, so many casinos only let new players join after a shuffle.

Back-counting deserves a number, because it is the method that works best on paper. A player who only joins at a true count of +1 or higher and uses a 1–12 spread wins 7.8 units per 100 hands actually played in our simulation, about five times the result of playing every hand. It also means leaving a table whenever the count drops, which is exactly the behaviour casinos look for.

Online blackjack

Counting does not work against computer blackjack (RNG games). The software reshuffles the virtual deck before every hand, so there is nothing to count. Live-dealer tables use a real shoe, but most of them cut it around the middle. At 50% penetration our simulation gives a counter almost nothing, and there is no way to sit out the bad counts either.

9. Is card counting legal?

Counting cards in your head is legal, because you only use information that every player at the table can see. What is illegal in many places is using a device to help you, such as a phone app or a hidden computer. In Nevada that is a criminal offence.

Being legal does not mean the casino has to let you play. In most places casinos are private businesses and may refuse your play or ban you from the tables. One well-known exception is Atlantic City: after a 1982 court case brought by the card counter Ken Uston, New Jersey casinos may not bar counters, but they are still allowed to shuffle early, add decks or limit bets.

10. Card counting myths

  • “You need to remember every card.” You only keep one number, the running count.
  • “You have to be a maths genius.” Counting is adding and subtracting 1. The hard part is speed, concentration and staying calm while losing.
  • “Counting means you win every session.” As shown above, about a third of perfect counters are still behind after 125 hours.
  • “Other players ruin the count.” Other players change which cards you get, but not what is left in the shoe. Their cards are as random as yours, and you count them too.
  • “Counting works online.” Not against RNG games, and hardly at live-dealer tables with a half-dealt shoe.

Frequently asked questions

How much can you win by counting cards?

In our simulation, a perfect counter with a $10 to $80 spread wins about $8 an hour on average in a good six-deck game (80 hands per hour). With a $10 to $120 spread it is about $12. Real players make mistakes and get shuffled out, so they earn less.

How long does it take to learn card counting?

Basic strategy and the Hi-Lo values can be learned in a few days. Counting fast and accurately under casino conditions usually takes weeks of daily practice, and the deviations come on top.

Which counting system should a beginner use?

Hi-Lo. More complex systems with values like +2 or side counts for Aces gain a little, but they are slower and lead to more mistakes, which cost more than the extra accuracy brings.

Does card counting work with 6 or 8 decks?

Yes, if the shoe is dealt deep enough. Our simulation used 6 decks and found a win rate of +0.96 units per 100 hands with a 1–8 spread at 75% penetration. Fewer decks make counting easier, and a shallow cut card makes it much harder.

How we simulated it

  • Simulator: our own program, written for this guide. In total 3.3 billion rounds, with 200 to 400 million rounds for each rule set and 2,000 careers of 200,000 hands for each betting strategy.
  • Rules: 6 decks, dealer stands on soft 17 (one variant: hits soft 17), dealer checks for blackjack, double on any two cards and after splits, split up to four hands, split Aces receive one card, late surrender, blackjack pays 3:2 (one variant: 6:5). Cut card after 75% (variants: 50%, 65%, 83%).
  • Player: one player alone at the table, total-dependent basic strategy, Hi-Lo running count updated with every visible card, true count from the exact number of unseen cards, the 22 deviations from section 5.
  • Check: with basic strategy alone our simulator gives a house edge of 0.36%. The Wizard of Odds lists about 0.33% for a six-deck game with these main rules. Our game does not allow resplitting Aces, which costs the player a little. For 6:5 blackjack we measured an extra 1.35 points; the Wizard gives 1.39.
  • Limits: perfect play is assumed, with no counting or betting errors. Casino countermeasures are not simulated. Real counters estimate the decks remaining to the nearest half deck, which costs a little. Back-counting assumes you can watch every round and join at any time.

Related guides

About the author: GamblingSociety is an independent site run by Andreas, who has operated casino and betting guides since the mid-2000s. No bookmaker or casino owns or pays for our content. More about us.