A variance swing is a short-term movement in actual results away from the mathematical result expected over a very large number of wagers. It is the bankroll rising or falling faster, farther or in a different direction than the long-run average suggests.
A player can be ahead in a negative-expectation game. A casino can lose on a game it normally wins. Neither result cancels the house edge. It shows that random outcomes create a wide range of short-term paths before the long-run average becomes easier to see.
The term in one sentence
Variance is the mathematical spread; a variance swing is the result you experience when that spread moves the bankroll.
The word “swing” is informal but useful. It describes the visible movement from one point to another: an upswing, a downswing, a sharp reversal or an unusually flat period.
What a swing is—and is not
| Term | Meaning | Example |
|---|---|---|
| Expected value | The long-run average result per wager | A bet loses 1 unit per 37 units wagered on average |
| Variance | A measure of how widely outcomes spread around the average | A game with many possible payouts has wider dispersion |
| Standard deviation | Variance expressed back in the original units | A practical scale for a typical range of movement |
| Variance swing | The actual short-run rise or fall produced by that randomness | A bankroll gains 20 units, then loses 35 |
| Streak | A run of similar outcomes | Eight losing hands in a row |
| Drawdown | The fall from a previous bankroll peak | From $1,500 down to $900 |
| Volatility | How sharply or frequently results tend to move | A jackpot slot is generally more volatile than an even-money table bet |
These terms overlap, but they are not interchangeable. A streak may contribute to a swing. A swing may create a drawdown. Volatility influences the typical size and shape of swings. Variance and standard deviation are the mathematical tools used to describe the spread behind them.
Why house edge does not predict tonight’s result
House edge answers a long-run pricing question: how much does the casino expect to retain per unit wagered, assuming the stated rules and strategy conditions? It does not say that every player will lose that percentage during one session.
Suppose a wager has a 2% house edge. The casino’s long-run expected win is $2 per $100 of action. A player who wagers $100 once does not usually lose exactly $2. The player may win the wager, lose all of it, push or receive one of several payouts. The 2% becomes meaningful as an average across a very large body of action.
The gap between the small expected edge and the much larger possible result on each wager is where swings come from.
A simple roulette calculation
Consider a one-unit red bet on a single-zero roulette wheel:
- win one unit with probability
18/37; - lose one unit with probability
19/37.
The expected result per spin is:
(18/37 × +1) + (19/37 × -1) = -1/37 ≈ -0.0270 units
So the expected loss is about 0.027 unit per spin, or 2.70 units over 100 spins.
But the standard deviation of one spin is almost one full unit. For 100 independent spins, the standard deviation of the total is approximately:
√100 × 0.9996 ≈ 10 units
That comparison matters:
- expected result after 100 spins: about -2.7 units;
- one-standard-deviation scale of the swing: about 10 units.
A wide range of winning and losing totals is therefore ordinary over 100 spins. The house edge is present, but the swing is much larger than the expected loss in a short sample.
This does not mean the edge is weak. It means short sessions are noisy.
The same edge can produce different swing profiles
Two games can have similar house edges and feel completely different because their payout distributions differ.
An even-money wager produces frequent small wins and losses. A jackpot slot may return many small or zero outcomes and concentrate a large part of its payback in rare prizes. Both can have a defined long-run return, but the second game can produce much sharper individual swings.
Bet structure matters inside the same game as well. In roulette, a straight-up number and an even-money bet share the same single-zero house edge under standard payouts, but the straight-up number wins much less often and pays much more when it hits. The expected percentage is the same; the variance is not.
For this reason, choosing a lower house edge does not automatically guarantee a smooth session. House edge describes price. Variance describes spread.
Player-side examples
A variance swing can appear as:
- losing ten blackjack hands in a short sequence despite using correct basic strategy;
- hitting a slot bonus early and remaining ahead for the rest of a brief session;
- missing many video-poker draws before completing a rare high-paying hand;
- winning several Banker bets in baccarat and then giving back the profit in a reversal;
- experiencing a long poker tournament downswing despite making reasonable decisions.
The correct interpretation is not that the game became due, cold, generous or hostile. It is that a random process produced one possible path through the distribution of results.
Casino-side examples
Casinos also experience variance. A high-limit baccarat player can win several large decisions and push the pit’s result negative for a shift. A table-games department may underperform theoretical win for a month. A slot floor can pay several major jackpots close together.
Operators therefore separate concepts such as:
- theoretical win: expected result from the action;
- actual win: what the casino really retained;
- hold percentage: actual win relative to a selected denominator, commonly drop for table games;
- variance: the statistical spread around the expectation;
- exposure: how much one player, table or event can move the result.
A negative result does not automatically prove weak controls, just as a strong result does not automatically prove excellent management. The operational question is whether the result fits plausible variance and whether ratings, procedures, limits, payouts and reporting were accurate.
Why volume usually smooths percentage results
As independent or weakly dependent wagering volume grows, random wins and losses tend to offset one another more effectively. The absolute dollar swing can still become large, but the result as a percentage of total action often becomes more stable.
This is why a casino with many tables and machines can usually absorb individual winning players better than a tiny operation with concentrated exposure. It is also why a player with a very small bankroll can go broke before long-run averages have any practical chance to emerge.
More volume does not remove randomness. It changes the scale at which the average becomes visible.
Measuring spread correctly
In statistics, variance is based on squared distances from the mean, while standard deviation is the square root of variance and returns the result to the original units. The NIST/SEMATECH engineering statistics handbook explains these measures of scale and notes that standard deviation restores the units that variance squares.
For casino analysis, standard deviation is often easier to communicate because the answer can be expressed in dollars, units or percentage points. The exact model still depends on the game, payout distribution, number of trials and independence assumptions.
Common mistakes when reading a swing
“The result must reverse soon”
Long-run convergence does not schedule a correction for the next hand. A downswing can continue, reverse immediately or move sideways.
“I am winning, so I found an edge”
A short upswing is compatible with both positive- and negative-expectation play. Skill or advantage requires evidence beyond one favorable result.
“The casino lost, so the game is vulnerable”
A negative shift can be ordinary variance. Control failures must be demonstrated through procedure, surveillance, equipment, rating or payout evidence.
“A smaller bet removes variance”
Smaller units reduce the dollar size of the swing relative to the bankroll. They do not eliminate the randomness of the outcomes.
“Variance is the same as risk of ruin”
Variance contributes to bankroll risk, but risk of ruin also depends on starting bankroll, bet size, edge, stopping rules and game structure.
Managing the effect, not predicting the path
A player cannot control which valid random outcome appears next. The controllable variables are bankroll, wager size, game choice, pace and stopping rules.
An operator cannot eliminate legitimate player wins. The controllable variables include table limits, credit, bankroll adequacy, game mix, concentration, rating quality and procedural protection.
That is the practical lesson of a variance swing: manage exposure so that an ordinary bad run does not force an extraordinary decision.
Related terms and tools
Continue with variance for the formal concept, standard deviation for the scale of spread, short-term variance for session interpretation and bankroll for the fall from a peak. The variance simulator can illustrate how many different result paths can emerge from the same assumptions.