Expected value, variance and standard deviation
Expected value (EV) describes the average net result in a specified model; variance describes how widely results fluctuate around that average. Standard deviation is the square root of variance and is expressed in the same units as the result. Variance itself uses squared units.
Why this matters when real money is involved
A small house edge is not a limit on session losses. Even with correct basic strategy, a short session can finish far above or below its expected result. A profitable session does not demonstrate a player advantage.
How results combine over many hands
In a simplified model of independent hands with the same fixed initial bet b, mean net return μ per initial betting unit and standard deviation σ per unit:
- Expected net result after N hands = N × b × μ.
- Variance of the total = N × b² × σ².
- Standard deviation of the total = b × σ × √N.
The modeled hand return must include any split and double outcomes on the stated initial-bet basis. If bets vary, hands are correlated or the strategy and rules change, these simple formulas need adjustment. Finite-shoe play is not literally a sequence of independent identical hands.
Winning and losing streaks
A streak alone is not evidence that the next wager is favorable. In a shoe, exposed card composition can change probabilities; merely labeling recent results “hot” or “cold” is not a substitute for that information. Raising stakes to recover losses increases exposure without establishing an edge.
A numerical illustration, not a prediction
Assume, solely for this example, μ = −0.005 units per hand (a 0.5% house edge), σ = 1.15 units per hand and a fixed $10 initial bet. The chosen σ is a model input, not a universal blackjack constant.
| Hands | Expected net result | Standard deviation of total |
|---|---|---|
| 100 | −$5 | $115 |
| 1,000 | −$50 | About $363.66 |
For 100 hands: 100 × $10 × (−0.005) = −$5, while $10 × 1.15 × √100 = $115. Standard deviation is neither a maximum loss nor an automatic confidence interval. This calculation does not give the probability of exhausting a bankroll before the final hand.
Budgeting for exposure
A budget limits what you are willing to spend; it does not change the game’s expected return. There is no universal number of betting units that guarantees surviving a session. Risk of ruin requires a defined bankroll, bet policy, time horizon and outcome model.
What basic strategy can and cannot do
Rule-matched decisions reduce avoidable expected loss. They do not remove uncertainty or promise income. Counting is a separate conditional analysis requiring suitable dealing conditions and accurate execution; it does not eliminate losing streaks either.
Changing bet size
With all other model inputs fixed, doubling every bet doubles the expected dollar loss and standard deviation, and multiplies variance by four. Reducing the stake reduces dollar exposure. Neither change reverses a negative expected return.
What the long run means
In the independent fixed-bet model, total standard deviation grows with √N, while expected loss grows with N. The average result per hand becomes less variable, but the expected cumulative loss still grows. Playing longer does not make a negative-edge game profitable or require a losing session to recover.
Keeping emotions separate from the model
Do not turn a statistical average into a target you must win back. Stop when your chosen limits are reached or when play becomes difficult to control. A break is an opportunity to stop exposure, not a method of resetting the cards.
Review the assumptions before using a result
Record the rules, bet size, number of hands and whether the estimate includes side bets, doubles and splits. Distinguish actual session records from theoretical calculations. A graph that looks smooth can still rest on inappropriate assumptions.
Useful tools for further study
Use the expected-loss calculator for average-cost scenarios and the risk-of-ruin simulator for its stated model. Read each tool’s limits; neither forecasts your next session.
Why the average bet can hide extra volatility
For predetermined, unequal stakes in the same independent-hand model, expected net result is μ × the sum of initial bets. Variance is σ² × the sum of squared initial bets. You cannot replace every stake with its average when calculating variance: larger bets receive more weight because the stakes are squared.
Using the illustrative inputs above, suppose five hands start at $10 and five at $20. Initial bets total $150, so expected net result is −0.005 × $150 = −$0.75. The sum of squared stakes is 5 × 10² + 5 × 20² = 2,500. Total standard deviation is 1.15 × √2,500 = $57.50.
Ten $15 bets also total $150 and have the same expected net result in this model, but their standard deviation is 1.15 × $15 × √10, about $54.55. The unequal schedule has greater volatility despite the identical average stake. This is a comparison of fixed hypothetical schedules, not a recommendation to increase stakes. Outcome-dependent bets, counting, shared dealer hands and stopping rules require a model of those dependencies.
One round can require more than the starting bet
A $10 initial wager is not necessarily the maximum cash committed to its round. If the rules allow four split hands and doubling all four, that round can require $10 × 4 × 2 = $80 before insurance or side bets. This is a conditional exposure example, not a statement that every game allows those actions. The per-initial-bet return model already includes permitted splits and doubles: do not add their risk a second time to that model. Separately check whether your remaining budget can cover the round before making its initial wager.
Mathematical references
The proof of variance additivity for independent variables explains the assumption behind combining hands. Michael Shackleford’s blackjack variance research shows how rules and shared dealer outcomes affect variance and covariance. Those game-specific values are not interchangeable with every table or the illustrative inputs used here.
Frequently asked questions
Can I lose more than the expected loss?
Yes. Expected loss is a model average, not a cap. A session may lose substantially more, finish near that average or win.
Does a larger bankroll reduce the house edge?
No. It can change the chance of exhausting the funds under a specified plan, but it does not change the rules or expected return per wager.
Compare results over a fixed number of hands
Session length must be stated before comparing outcomes. A result after 100 hands and a result after 10,000 hands come from different exposure. Record hands played, the size of each initial wager, rules, strategy, and any bet changes. These inputs make the expected value and standard-deviation estimates reproducible and prevent a short winning run from being mistaken for a reliable long-term rate.
