SLO 115: Slot Variance: What the Number Actually Measures

A slot can have a 96% RTP and still produce either a fairly steady sequence of small returns or long empty stretches interrupted by rare large awards. RTP alone cannot describe that difference. Variance can.

Slot variance is a mathematical measure of how widely the possible net results of one wager are spread around the expected result. It is not the difference between what a player expected to lose and what happened in one session. It is a property of the game's complete probability-and-payout distribution.

That distinction matters because a short session does not reveal the machine's true variance reliably. The result can be extreme precisely because the variance is high.

Start with the payout distribution

Assume a simplified $1 slot has these possible returns:

The expected net result is:

Expected Net Result = Σ(Probability × Net Result)
                    = (0.70 × -$1) + (0.20 × $0) + (0.08 × $1) + (0.02 × $29)
                    = -$0.04

The game therefore returns an average of $0.96 per $1 wagered, which is a 96% RTP and a 4% house edge.

The average does not show the shape. Most spins lose the entire wager, while a small part of the return sits in the $30 outcome. Variance measures that unevenness.

The variance calculation

For a population of known outcomes, variance is:

Variance = Σ[pᵢ × (xᵢ - μ)²]

Where:

• pᵢ is the probability of outcome i;
• xᵢ is the net result of that outcome;
• μ is the expected net result;
• Σ means add the result for every possible outcome.

Using μ = -$0.04, the simplified game has:

Variance = 0.70(-1 + 0.04)²
         + 0.20(0 + 0.04)²
         + 0.08(1 + 0.04)²
         + 0.02(29 + 0.04)²
         = 17.5984 dollars²

Variance is expressed in squared units, which is mathematically useful but awkward to interpret. Taking the square root gives the standard deviation:

Standard Deviation = √17.5984 = about $4.20 per spin

Standard deviation restores the result to dollars. NIST's explanation of variance and standard deviation as measures of spread explains why the square root is commonly used when interpreting dispersion.

Same RTP, completely different ride

Now compare the first game with a smoother $1 game:

Both games lose an average of four cents per $1 wager in the long run. The second places far more weight in a rare large outcome, so its result distribution is much wider.

This is why “choose the highest RTP” is incomplete advice. RTP identifies the long-run center. Variance helps describe the distance a real session may travel from that center.

Variance, volatility, and hit frequency are related but not interchangeable

Casinos, manufacturers, analysts, and players sometimes use variance and volatility loosely. A useful separation is:

• Variance is the calculated spread of the outcome distribution.
• Standard deviation is the square root of variance and is easier to express in wager units or money.
• Volatility is the broader practical description of how violently results tend to move. It may be communicated as low, medium, or high rather than as a published number.
• Hit frequency is the probability that a spin returns any amount defined as a hit. It says nothing by itself about the size of those returns.

A game can have frequent small hits and still be high variance if enough of its return is concentrated in a rare jackpot. It can also have a low hit frequency without an enormous top prize. The complete pay distribution decides the variance.

For the player-facing interpretation, see Slot Volatility Explained. For the separate question of how often a listed event occurs, see Slot Hit Frequency.

How session uncertainty scales

If wagers are independent and the distribution stays the same, the standard deviation of the sum of n wagers is approximately:

Session Standard Deviation = Per-Wager Standard Deviation × √n

For the spiky example, 400 $1 spins produce:

Expected Session Result = 400 × -$0.04 = -$16
Session Standard Deviation = $4.20 × √400 = about $84

This does not mean the session must finish between -$100 and +$68. Standard deviation is not a hard boundary, and a slot distribution with rare large awards may be skewed rather than normally distributed. The calculation simply shows that a result tens of dollars away from the $16 expected loss is not surprising.

The formula also explains a common misconception. More spins make the average return per spin more stable relative to the amount wagered, but the absolute dollar swing of the total result can still grow.

What casinos can and cannot infer from actual results

Slot departments compare actual hold with theoretical hold over controlled reporting periods. A large gap can come from ordinary variance, an incorrect theoretical setting in the report, meter or configuration problems, jackpot timing, excluded promotional transactions, or other data-quality issues.

A sound review asks:

1. Is coin-in complete and assigned to the correct machine and paytable?
2. Is the theoretical hold percentage current?
3. Were jackpots, fills, promotional awards, and machine moves treated consistently?
4. Is the sample large enough for the game's expected fluctuation?
5. Does the difference persist across rolling periods?

Variance is not a reason to ignore anomalies. It is a reason to test them against the game's expected distribution instead of reacting to one lucky or unlucky weekend.

What players cannot learn from a short session

A few hundred spins cannot establish that a machine is loose, tight, due, altered, or paying below its approved return. A personal record contains too little information to recover the full hidden distribution, especially when rare features carry substantial value.

Session observations can answer personal questions: how much was wagered, what was won, how quickly the balance moved, and whether the game fits the player's loss limit. They usually cannot answer the design question, “What is this game's true variance?”

The Variance Simulator can demonstrate how different paths emerge from the same expectation, but a simulation is only as accurate as the probabilities entered.

Practical reading of the number

For a player, higher variance generally means a greater chance of sharp drawdowns and a stronger dependence on rare awards. Lower variance generally means smaller changes per wager, not a positive expectation. A low-variance game with a house edge can grind down a bankroll steadily.

For an analyst, variance should be tied to the exact wager unit and paytable. Multiplying every wager by five multiplies standard deviation by five and variance by twenty-five. Comparing raw variance figures without normalizing the stake is meaningless.

For an operator, theoretical return and variance answer different questions:

• RTP: What is the long-run average return?
• Variance: How widely can outcomes spread around that average?
• Actual hold: What happened in the measured period?
• Standard deviation or confidence model: Is that result unusual for the play volume and distribution?

Read Slot RTP, Standard Deviation, and Expected Value for the supporting mathematics.

The precise takeaway

Variance is not casino jargon for “unpredictable.” It is a calculation built from every possible net result and its probability. It explains how two games with identical RTP can create very different bankroll paths. It does not forecast the next spin, prove a machine's configuration from a small sample, or make a negative-expectation game safe.