Variance explains why gambling results swing wildly in the short term. This guide covers how variance, standard deviation, and risk of ruin affect your bankroll, why streaks happen even in low-edge games, and how mathematical bankroll strategies help manage volatility.
A casino game can have a small house edge and still produce long losing streaks. A blackjack table with a 0.5% house advantage can generate sessions where a player loses 20 or 30 bets in a row. This occurs because gambling outcomes depend not only on expected value (EV) but also on variance. Variance determines how widely actual results fluctuate around the expected outcome.
Understanding variance explains why short sessions often contradict the long-run math. It also explains why bankroll size matters and why even players with positive expectation can go broke. This guide examines variance, introduces standard deviation, explains the risk of ruin formula, and shows how bankroll size interacts with gambling volatility.
In statistics, variance measures how far outcomes deviate from the expected value.
The mathematical definition is:
$$Variance = E[(X – \mu)^2]$$
Where:
Variance measures the average squared distance from the mean outcome.
For gambling analysis, variance describes how large the swings around expected results can become.
Example:
Suppose a game has an expected loss of €1 per round.
Possible outcomes might be:
| Outcome | Probability |
|---|---|
| Win €10 | 10% |
| Lose €1 | 90% |
Even though the average result approaches −€1, individual results vary widely.
Variance therefore measures how violent the swings around the average can be.
Variance and expected value (EV) describe two completely different properties of a game.
EV measures the long-run average result.
The EV formula is:
$$EV = p \times w – (1 – p) \times l$$
Where:
Example:
$$EV = 0.49 \times 1 – 0.51 \times 1$$
$$EV = -0.02$$
Result:
Expected loss of 2 cents per €1 bet.
Variance, however, measures how far actual outcomes may stray from that average.
Two games may have identical EV but widely different variance.
Example:
| Game | EV | Variance |
|---|---|---|
| Roulette | −2.7% | Medium |
| Slots | −4% | Very high |
Slots often have larger payouts and longer losing streaks, which increases variance.
The relationship between EV and house edge is covered here:
Variance arises because casino outcomes are random events.
Short sessions rarely match the theoretical average.
Consider a roulette red/black bet.
Probability of losing:
$$P(loss) = \frac{19}{37}$$
Convert to decimal:
$$P(loss) = 0.5135$$
Even though this probability is close to 50%, the probability of several consecutive losses remains significant.
Probability of 5 losses in a row:
$$P = 0.5135^5$$
Step calculation:
$$0.5135 \times 0.5135 = 0.2637$$
$$0.2637 \times 0.5135 = 0.1354$$
$$0.1354 \times 0.5135 = 0.0695$$
$$0.0695 \times 0.5135 = 0.0357$$
Result:
$$3.57\%$$
This explains why streaks are common even in nearly even-chance games.
Casino math assumes a large number of trials.
The Law of Large Numbers states that the average outcome moves toward expected value as sample size increases.
However, short sessions involve small sample sizes where variance dominates.
Example:
Expected loss with 4% house edge:
$$Expected\ Loss = Total\ Wagered \times House\ Edge$$
If €1 is bet on 1,000 spins:
$$Expected\ Loss = 1000 \times 0.04$$
$$Expected\ Loss = 40$$
But a short 100-spin session could produce results anywhere between −€150 and +€100 depending on variance.
Variance allows temporary results far from expectation. Even players with positive EV can lose large amounts in the short term.
Example:
A blackjack card counter may have:
| Metric | Value |
|---|---|
| Player edge | +1% |
| Bet size | €100 |
| Hands played | 100 |
Expected profit:
$$EV = €100 \times 100 \times 0.01$$
$$EV = €100$$
But the standard deviation per hand in blackjack is roughly 1.15 units.
That produces swings much larger than the expected profit.
Even players with an advantage therefore need large bankrolls.
Variance itself is difficult to interpret because it uses squared units.
Instead, analysts use standard deviation $(SD)$.
Standard deviation equals the square root of variance.
$$SD = \sqrt{Variance}$$
For repeated trials, total standard deviation grows with the square root of the number of rounds.
$$SD_{total} = SD_{per\ round} \times \sqrt{n}$$
Where:
Example:
If SD per round = 1 unit and 100 rounds are played:
$$SD = 1 \times \sqrt{100}$$
$$SD = 1 \times 10$$
$$SD = 10$$
This means results after 100 rounds tend to fluctuate ±10 units around EV.
Approximate standard deviation values vary widely between games.
| Game | Approx SD per round |
|---|---|
| Blackjack | ~1.15 units |
| Roulette (even bet) | ~1 unit |
| Baccarat | ~1 unit |
| Video Poker | 4–5 units |
| Slots | 5–20+ units |
Slot machines often have extreme variance due to jackpot payouts.
Variance introduces the possibility that a bankroll drains before expected value appears.
Gamblers call this the Risk of Ruin (RoR).
One simplified form of the RoR formula is:
$$RoR = \left(\frac{1 – e}{1 + e}\right)^{B/u}$$
Where:
Example:
Assume:
Player edge:
$$e = 0.02$$
Bankroll:
$$B = 1000$$
Bet size:
$$u = 10$$
Calculate exponent:
$$B/u = 100$$
Substitute:
$$RoR = \left(\frac{1 – 0.02}{1 + 0.02}\right)^{100}$$
$$RoR = (0.98 / 1.02)^{100}$$
$$RoR = (0.9608)^{100}$$
Approximate result:
$$RoR \approx 1.7\%$$
Meaning there is roughly a 1.7% chance the bankroll goes to zero before the edge appears.
Slot variance uses a volatility index.
High volatility slots produce large jackpots, long losing streaks, and high standard deviation. Low volatility slots produce frequent small wins and a smoother bankroll curve.
Blackjack has moderate variance because outcomes are close to even money, payouts are consistent, and blackjack bonus payouts add some additional swing. Standard deviation is roughly 1.15 betting units per hand.
Roulette variance depends on bet type.
| Bet Type | Variance |
|---|---|
| Red/Black | Low |
| Dozen | Medium |
| Straight number | High |
Straight bets create higher variance because payouts are 35:1.
Video poker variance varies by paytable.
| Game | Variance |
|---|---|
| Jacks or Better | ~19 |
| Double Double Bonus | ~42 |
Higher variance versions require larger bankrolls.
The Kelly Criterion determines optimal bet size for positive-EV bets.
Formula:
$$f^* = \frac{bp – q}{b}$$
Where:
Example:
Even-money bet:
$$b = 1$$
Probability:
$$p = 0.55$$
$$q = 0.45$$
Calculation:
$$f^* = \frac{1 \times 0.55 – 0.45}{1}$$
$$f^* = 0.10$$
Result:
Bet 10% of bankroll.
In casino games with negative EV, Kelly produces 0% bet size, meaning the optimal wager is not to play.
Variance determines how large a bankroll must be relative to bet size.
Typical bankroll guidelines:
| Game | Suggested Bankroll |
|---|---|
| Blackjack | 100–200 units |
| Roulette | 100 units |
| Video poker | 300–500 units |
| High-variance slots | 500+ units |
These values aim to keep risk of ruin low.
Monte Carlo simulations model thousands of random gambling sessions. Many simulated paths lose heavily early. Some paths show large profits. Most paths eventually converge toward expected value. Simulations show how variance dominates short sessions while EV dominates very large sample sizes.
Variance measures how widely results fluctuate around expected value. High variance produces larger bankroll swings.
Risk of ruin uses formulas that include bankroll size, bet size, and player edge.
Safe bankroll sizes range from 100 to 500 betting units depending on game variance.
High variance increases the probability of large wins but also increases the probability of large losses.
Standard deviation measures the typical size of deviations from expected value.
Betting systems do not change the probability distribution of outcomes, so variance stays the same.
Variance can produce long losing streaks that drain a bankroll before the statistical edge appears.
House edge measures the long-run expected loss, while variance measures how volatile the short-term results can be.