Hi-Lo Counting System

Overview

Hi-Lo (also written Hi-Lo or High-Low) is a level-one balanced counting system. “Level one” means every card is tagged with only −1, 0, or +1 — never ±2 or fractions. “Balanced” means that if you count through an entire deck (or shoe) without error, the running count finishes at zero.

That balance property is powerful: it lets you convert the running count into a true count (count per remaining deck), which normalizes the advantage across single-deck, double-deck, and multi-deck shoes. Almost every modern training program and casino-simulation study uses Hi-Lo as the baseline comparison system.

Historically, the core idea descends from Edward Thorp’s work in the 1960s and was refined into the simple high-card / low-card dichotomy that players use today. Writers such as Stanford Wong popularized practical Hi-Lo procedures for casino conditions: how to size bets, when to take insurance, and how to “Wong in” (enter a table only when the count is favorable).

Approximate efficiency figures commonly cited for Hi-Lo; exact values depend on rules and simulation methodology. Use them comparatively, not as guarantees.

Tag values

Every card you see gets a value. You do not need to remember suits — only ranks. The intuition is simple: low cards help the dealer more when removed (they “bust-proof” stiff hands), so when lows leave the shoe the remaining cards are richer in tens and aces for the player. High cards help the player (more blackjacks, better double-downs, dealer busts more often), so when highs leave, the shoe gets worse for you.

Running count

The running count (RC) starts at 0 after a shuffle (for a balanced system). As each card is exposed — player hands, dealer upcard, hole card when revealed, burned cards if visible — you add its tag to the running total. You keep this number continuously, not just once per round.

A rising RC means more low cards have been seen than high cards → the remaining shoe is rich in tens and aces → player edge improves. A falling RC means the opposite: high cards are depleted and you should bet small or leave.

What to count

• All player cards that are face-up.
• Dealer’s upcard immediately.
• Dealer’s hole card when it is flipped (after players finish, or immediately on blackjack checks depending on procedure).
• Any cards dealt face-up in multi-hand or side spots you can clearly see.

You do not need the value of face-down cards you never see. Perfect information is not required — card counting estimates composition from the cards that are exposed.

True count

In multi-deck games, a running count of +6 with six decks still in the shoe is weak; the same +6 with only two decks left is strong. True count (TC) fixes this:

Estimating decks remaining

Glance at the discard tray. If a six-deck shoe started full and roughly two decks are discarded, about four decks remain. With practice you estimate to the half-deck. Error here is a major source of real-world disadvantage versus perfect simulation — train it deliberately.

Examples

• RC = +8, decks left ≈ 4 → TC ≈ +2
• RC = +8, decks left ≈ 2 → TC ≈ +4
• RC = −6, decks left ≈ 3 → TC ≈ −2

Bet size and many strategy deviations key off true count, not raw running count (except in single-deck games where RC and TC are nearly the same early in the deck).

Why it works

Blackjack is not an independent trial like roulette. Cards are dealt without replacement until a shuffle. Removing cards changes the probabilities of:

• Player blackjack (3:2 or 6:5 payouts make ten+ace especially valuable)
• Successful double-downs (more tens improve many doubles)
• Dealer busts on stiff upcards (2–6) when the shoe is ten-rich
• Insurance (pure ten-density proposition)

Hi-Lo’s tags are a compact statistic for “how ten/ace-rich is the remainder?” When TC is high enough, the player’s expected value can flip from slightly negative (the house edge under basic strategy) to slightly positive — enough that larger bets in those moments can overcome smaller bets when the shoe is cold.

Efficiency metrics

Three classic metrics compare counting systems:

• Betting correlation (BC) — how well the count tracks advantage for bet sizing. Hi-Lo scores very high (~0.97).
• Playing efficiency (PE) — how well it guides insurance and strategy deviations. Hi-Lo is moderate (~0.51); specialized systems score higher.
• Insurance correlation (IC) — alignment with ten density for insurance decisions (~0.76 for Hi-Lo).

For most recreational and semi-serious players, Hi-Lo’s excellent BC is the priority: most of the gain from counting comes from betting more when you have the edge, not from the smaller gains of perfect index plays. Advanced systems improve PE at the cost of mental load.

Betting strategy

A betting ramp maps true count to unit size. Exact numbers depend on bankroll, risk of ruin, and table minimum/maximum. Conceptually:

Playing decisions & index plays

Basic strategy is still the foundation. Counting adds index plays (Illustrious 18 / Fab 4 style lists are often paired with Hi-Lo): situations where you deviate when the true count crosses a threshold.

High-value Hi-Lo style deviations (illustrative)

• Insurance — take when TC is high enough (often cited near +3 in multi-deck; confirm for your rule set).
• 16 vs 10 — stand at higher counts instead of hitting.
• 15 vs 10 — stand at sufficiently high TC.
• 10 vs 10 / 10 vs A — double more aggressively at high TC; avoid at low TC.
• 12 vs 3, 12 vs 2 — stand earlier when counts rise (dealer more likely to bust).
• Late surrender indices when the game offers it.

Full index tables fill pages and depend on decks and rules (H17 vs S17, DAS, surrender, etc.). Train basic strategy to automatic first, then add the most valuable indices one cluster at a time.

Insurance in depth

Insurance is a side bet that the dealer has blackjack. It pays 2:1 and is actuarially poor with a full shoe — but it becomes correct when ten-density is high enough. Because Hi-Lo tracks tens and aces together (aces also −1), it is a good but not perfect insurance predictor; systems with separate ace side counts can refine this further.

In practice, many Hi-Lo players use a simple true-count cutoff for insurance and ignore finer adjustments until their primary count is flawless.

Worked example

Pros & cons

Training plan

1. Week 1 — Tags only. Drill flash cards until each rank maps instantly. Count down a single deck to 0 in under 30–25 seconds.
2. Week 2 — Running count. Deal hands face-up; keep RC without pausing. Aim for zero errors over 2–3 shoes.
3. Week 3 — True count. Estimate discard tray every round; convert RC → TC aloud, then silently.
4. Week 4 — Betting ramp. Assign chips to TC; practice smooth spreads without telegraphing.
5. Ongoing — Indices. Add insurance first, then Illustrious-18 style plays. Re-test full-deck accuracy weekly.

Performance targets used by many trainers: near-100% tag accuracy, single-deck count-down under ~20–25 seconds, and multi-deck simulation accuracy above ~98% before considering live play.

Common mistakes

• Skipping true count in multi-deck games — RC alone mis-sizes bets badly.
• Miscounting aces as zero — in Hi-Lo aces are −1, not neutral.
• Updating only once per round — update as cards appear so mid-round decisions (insurance, doubles) use current info.
• Ignoring shuffle points — know pen (how deep the shoe goes) and reset RC only on shuffle.
• Chasing losses with the count — TC dictates bet size; emotion does not.
• Over-focusing on rare indices before the running count is automatic.