Card Counting High Low System: The 5 Parts That Win Money

The card counting high low system wins money because it converts a guess about deck composition into a bet size, and the math behind that conversion has been settled since Harvey Dubner published it in 1963. The system has five connected parts: the card values, the running count, the true count conversion, the bet ramp, and the I18 plus Fab 4 deviations. Each part earns a specific share of the edge. Skip any of them and you are paying full house edge with a side of mental tax.

This post is not the canonical Hi-Lo explainer. That one already lives at card counting high low. This one is the architecture. Why each piece is shaped the way it is, what each one earns, and what happens when the part breaks. If you have already memorized the values and you still cannot understand why your sessions are not matching the math, this is for you.

Card counting is just math. The casino has known about it since 1963. They are still in business.

How the card counting high low system actually wins money

Hi-Lo is a balanced level-1 system. Three card values, one running total, one division step, one bet ramp, twenty-two deviation plays. That is the whole stack. Sixty-three years of math and millions of hours of live play have not produced a better practical system for shoe blackjack.

The base house edge against perfect basic strategy at a 6D H17 DAS LS table is -0.47%. A clean Hi-Lo player running a 1-8 spread at decent penetration pulls to roughly +1.0 to +1.5 percent edge depending on ramp shape. That swing from -0.47 percent to +1.0 percent is the entire game. Everything else in the system exists to produce that swing.

The five parts of the system map to five distinct jobs. Values convert the cards to numbers. The running count tracks the deck composition. The true count normalizes for shoe depth. The bet ramp converts the count to a bet. The deviations convert the count to a play. Break any single mechanism and the whole system loses its edge. A counter with perfect values and a perfect running count who flat-bets every hand has the exact same expected loss as a tourist who never heard of Hi-Lo. The architecture only works as a unit.

Hi-Lo is still the right starting system for 95 percent of counters. A clean Hi-Lo player captures roughly 97 percent of the theoretical edge available to a perfect Wong Halves player. The remaining 3 percent gets eaten three times over by the higher error rate of level-2 and level-3 systems under casino conditions. The fancy systems are a hobby for people who already mastered Hi-Lo and got bored. Working pros stay on Hi-Lo because Hi-Lo wins money.

Card counting values: why +1, 0, -1 are exactly the right numbers

The card counting values fit on three lines:

• Cards 2, 3, 4, 5, 6: +1
• Cards 7, 8, 9: 0
• Cards 10, J, Q, K, A: -1

The buckets are not arbitrary. They follow directly from how blackjack rules favor each side at different shoe compositions.

Low cards help the dealer. A dealer holding a stiff hand of 12, 13, 14, 15, or 16 has to hit. Low cards in the remaining shoe let the dealer make 17 to 21 out of those hands. When the 2s through 6s leave the shoe, the dealer busts more often, and the player wins more pushes against bad dealer up-cards. That is worth a +1 on each low card.

High cards help the player. The player gets paid 3:2 on a natural blackjack. The dealer gets paid 1:1 on the same hand. Tens and aces make blackjacks. When those cards remain in the shoe, the player wins bigger on the hands they win. Double downs also hit harder when the next card is likely to be a ten. That is worth a -1 on each high card.

The 7s, 8s, and 9s are roughly neutral by the math because they help both sides about equally. They make some dealer stiffs and break some player double downs. The net effect washes out. So they earn a 0.

Why not +2 on a 5 and +1 on a 2? Because the index gain on the higher-weighted values is real but tiny, and the error rate at the table from tracking multiple weights at speed is huge. Hi-Lo trades a small theoretical loss for a large practical gain. That trade is the entire reason it became the standard.

The running count: what every card moves and why

The running count is one number that updates with every card dealt. Start at zero. Add or subtract the value of each card as it leaves the shoe. The number you are holding at any moment is the running count.

The reason it works is that the running count is a one-dimensional projection of a high-dimensional reality. There are 312 cards in a 6-deck shoe. The state of the shoe at any moment is a 312-element vector. Tracking that vector is impossible at speed. The Hi-Lo bucketing collapses it to a single integer that still preserves enough information to size a bet correctly.

The check that the system is internally consistent: a balanced count means the values across a fresh shoe sum to zero. Six decks of 52 cards is 312 cards. Twenty cards (2 through 6 of each suit, four times) count +1 each. Twenty cards (10, J, Q, K, A of each suit, four times) count -1 each. The 12 zero-value cards in each deck contribute nothing. Over a full shoe the running count starts at zero and returns to zero. If you drill a single deck and end at something other than zero, you missed a card. That self-correction is built into the system on purpose.

Drill the deck count until it is automatic before drilling shoes. Single deck face-up, end at zero, under 30 seconds. A counter who has not nailed the single-deck count is going to leak running count errors at the table. The closer reads on the math behind each card counting values bucket are worth understanding, but the drilling is non-negotiable.

The true count: why you divide by decks remaining

The running count tells you what cards have already been dealt. The true count tells you what is left in the remaining shoe. They are different numbers, and the bet ramp keys off the second one.

The formula:

True Count = floor(Running Count / Decks Remaining)

A running count of +6 with two decks left in the shoe is a true count of +3. A running count of +6 with four decks left is a true count of +1. Same running count, different bet entirely.

Why the division? Because the edge from card counting scales with the concentration of high cards left, not the absolute number. A running count of +6 with two decks remaining means three extra high cards per deck. A running count of +6 with four decks remaining means one and a half extra high cards per deck. The first shoe is twice as profitable on the next hand as the second one. The math demands the normalization.

Always floor the division. Never round. A running count of +5 with two decks left is TC+2, not TC+3. The half-deck rounding error compounds over a thousand hours into real money walking off the table. The system was designed around the floor. Skipping it is breaking the architecture.

Estimating decks remaining is the other half of the conversion. Look at the discard rack. With practice you can call a 6-deck discard stack within a half deck just by eyeballing it. The discard rack is usually marked on the side at half-deck increments. Drill discard-rack estimation as a separate skill from the running count. They are two muscles, not one.

The true count is the trigger for the bet ramp and the deviations. Everything keys off the true count. Getting this conversion wrong is the most costly single error in the system. The deeper read on the math behind the true count walks through worked examples.

The bet ramp: where the entire edge actually lives

The bet ramp is the conversion step from count to money. A counter who runs a perfect Hi-Lo count and flat-bets every hand has the exact same expected loss as a non-counter, plus a tax on attention. The count is the trigger. The spread is the engine.

A realistic 1-8 spread at a 6D H17 DAS LS game with a $10 minimum looks like this:

• TC 0: $10
• TC+1: $30
• TC+2: $60 (or $30 across two hands)
• TC+3: $100
• TC+4: $200
• TC+5: $300
• TC+6 and above: $600

The ramp shape matters as much as the spread number. Linear ramps draw heat. Stepped ramps with the occasional flat round at TC+1 read more like a player who got lucky than a working AP. Cover is part of the system, not a separate skill bolted on the side.

The hard part of the ramp is not the math. The math is learnable. The hard part is putting the big bet out when the count calls for it.

Early on I convinced myself the entire casino was watching the moment I pushed $300 across two hands. So I'd shade the bet down. Put out $100 instead of $300. Tell myself I'd spread harder next shoe. The reality: 90 percent of the time the pit is watching the dealer, not the players. Their job is making sure the house does not make a payout error. You are not the priority you think you are. The cost of shading the ramp at TC+4 is the entire reason you sat down. The count earned that bet. Refusing to take it is leaving EV on the floor every shoe.

Most counters never get reviewed by another player. That is the gap. Self-grading a bet ramp from inside the session almost always undersizes the bet. The data shows what the session log cannot.

If you are scared to spread, do not count yet. Build the bankroll, drill the ramp at table pace, and come back when the bet feels neutral. The discipline to push the chips out is the only edge that matters once the math is settled.

Illustrious 18 and Fab 4: deviations that earn the math

Deviations are basic strategy plays that change based on the true count. A standard basic strategy player always hits 16 versus 10. A Hi-Lo counter stands on 16 versus 10 at TC 0 and above. The play changes because the deck composition changes the math underneath it.

The Illustrious 18 is Don Schlesinger's list of the 18 highest-EV index plays. The Fab 4 is his list of the four surrender index plays. Together they cover 22 indices. Schlesinger demonstrated that these 22 plays capture roughly 80 percent of the available index EV. Drilling beyond the I18 plus Fab 4 hits diminishing returns fast.

The most valuable single deviation is insurance at TC+3. Insurance is normally a -7 percent house-edge sucker bet. At TC+3 and above the math flips and insurance becomes a positive-EV side bet on every dealer ace. A Hi-Lo player who takes insurance at TC+3 and refuses it below picks up meaningful EV per hour at heavy spreads. The deeper read on insurance in blackjack covers the conversion point and the rule of thumb.

The other top-EV I18 plays:

• 16 vs 10 at TC 0+: stand (basic strategy says hit)
• 15 vs 10 at TC+4: stand
• 12 vs 3 at TC+2: stand
• 12 vs 2 at TC+3: stand
• 10,10 vs 5 at TC+5: split (this one looks crazy to the table)
• 9 vs 2 at TC+1: double
• 11 vs A at TC+1: double

Drill the play at the count where it triggers, not as flashcards in isolation. The transition you want is from "what does the chart say" to "the count is plus four, I have 15 versus 10, stand." That is what shoe-mode practice trains. Index accuracy at high counts beats basic strategy purity every time, and skipping the deviations is paying for the count without using it.

What breaks the Hi-Lo system for most counters

The system breaks at predictable seams. Every plateaued counter is failing at one of these in a way they cannot see from inside the game.

Counting at the wrong game. A 6:5 blackjack payout adds 1.39 percent to the house edge and erases any count edge a 1-12 spread could ever produce. Walk past 6:5 tables. There is no recoverable EV at any spread. The same is true of continuous shuffle machines, which eliminate the count entirely.

Skipping basic strategy. A counter who is sloppy on basic strategy is donating a bigger edge than they will ever recover by counting. Drill basic strategy until automatic before adding the count. The order is not negotiable.

Undersizing at high counts. The mathematical bet ramp says push the money out at TC+4. The actual bet for most self-taught counters at TC+4 is half what it should be. Fear of attention, fear of losing, fear of the spread. The data from a logged session is the only way to see it. Most counters never see it.

Skipping deviations. Hitting 16 versus 10 at TC+4 is paying for the count without using it. The I18 plus Fab 4 is where the count earns most of its money in shoe games.

Tipping at high counts. A $10 tip at TC+4 in a six-deck shoe is real EV walking off the table. Tip at low counts if at all.

Variance does not care that you did everything right. A skilled Hi-Lo counter loses about four sessions out of ten. The graph dips and turns like the stock market and goes in one direction over time. The losing sessions are the default, not an anomaly. Most aspiring counters quit on session four, decide the system is broken, and walk. The system is not broken. They are looking at a sample size too small to mean anything.

The card counting high low system has not changed since 1963. The work to make it earn has not changed either. The free CountEdge trainer drills the running count, true count, I18, Fab 4, and bet ramp under shoe conditions with no credit card. Roughly four months at 20 focused minutes a day is what most counters need to get to a working table game. The longer version of why the app exists, and why a real working counter built it, is on the about page. The math is settled. The discipline is what is left.