St · Statistics & Probability

What Is the Mathematical Model of a Slot Machine?

Published By the 9to5Science editorial team

At a glance

  1. 1. List outcomes

    Represent each possible result of a spin and assign it a probability.

  2. 2. Apply the paytable

    Pair each result with its payout, using a consistent convention for whether payouts include the stake.

  3. 3. Calculate averages

    Use the probabilities and payouts to find the hit rate and expected return.

  4. 4. Measure variability

    Use the payout distribution to describe how much individual spins can differ from the average.

A slot-machine spin is a random experiment: it produces an outcome, and the paytable assigns that outcome a payout. Once you know the probability of each outcome and its payout, you can calculate how often prizes occur and what the game returns on average.

The exact probabilities and paytable vary by game. A simple model is useful for understanding the mathematics, but it does not predict the result of a particular spin.

Represent a spin as a probability distribution

Let the possible outcomes be o1,o2,…,ono_1,o_2,\ldots,o_n. Assign each outcome a probability pip_i, where pi≥0p_i\geq 0 and

∑i=1npi=1.\sum_{i=1}^n p_i=1.

Let rir_i be the payout for outcome oio_i. Choose a convention and stick with it. Here, a payout is the gross amount returned, including any stake returned. For a one-unit bet, a payout of 1 unit means the player gets the stake back and breaks even; a payout of 0 means the stake is lost.

The payout per spin is a random variable RR. Its expected value is

E[R]=∑i=1npiri.\mathbb{E}[R]=\sum_{i=1}^n p_i r_i.

If the stake is bb units, the net result is X=R−bX=R-b. Thus,

E[X]=E[R]−b.\mathbb{E}[X]=\mathbb{E}[R]-b.

This average does not say what happens on a particular spin. A player may win or lose much more than the average over a short run.

Model reels with symbol probabilities

For a simplified machine with independent reels, assign a probability to each symbol on each reel. If a target combination requires symbol AA on reel 1, BB on reel 2, and CC on reel 3, then

P(A,B,C)=P1(A)P2(B)P3(C),P(A,B,C)=P_1(A)P_2(B)P_3(C),

provided the reel outcomes are independent. If all reels have the same symbol probabilities, this becomes the product of those probabilities.

This is a mathematical model, not a requirement that a physical or digital slot literally use equally likely symbols. The symbols can have unequal probabilities. In real games, a random-number generator and game rules determine the outcome probabilities; a visible reel image alone may not reveal them. Bonus rounds or progressive features may also need to be modeled as additional outcomes or states.

Worked example: three independent reels

Suppose each of three reels independently shows AA with probability 0.50.5 and a non-AA symbol with probability 0.50.5. Consider this one-unit paytable:

  • AAAAAA pays 4 units gross.
  • Exactly two AA symbols pay 1 unit gross.
  • All other results pay 0.

The probability of AAAAAA is

P(AAA)=0.53=0.125.P(AAA)=0.5^3=0.125.

There are three ways to get exactly two AA symbols: AANAAN, ANAANA, and NAANAA, where NN means any non-AA symbol. Each has probability 0.530.5^3, so

P(exactly two A)=3(0.53)=0.375.P(\text{exactly two }A)=3(0.5^3)=0.375.

The remaining probability is 1−0.125−0.375=0.51-0.125-0.375=0.5. The expected gross payout is therefore

E[R]=0.125(4)+0.375(1)+0.5(0)=0.875.\mathbb{E}[R]=0.125(4)+0.375(1)+0.5(0)=0.875.

For a one-unit stake, the expected net result is 0.875−1=−0.1250.875-1=-0.125 units per spin. The return-to-player percentage (RTP) in this model is 0.875/1=87.5%0.875/1=87.5\%, and the corresponding house edge is 12.5%12.5\%. These long-run averages do not guarantee that a player will lose exactly 12.5 cents per one-unit spin over a short session.

The chance of receiving any payout is 0.125+0.375=0.50.125+0.375=0.5, or 50%. But only the AAAAAA result yields a net win: a one-unit gross payout merely returns the one-unit stake. This illustrates why “hit frequency” needs a definition; it often means the chance of any listed prize, including prizes that do not produce a net profit.

Measure payout variability

Two games can have the same expected return but very different swings. One way to quantify variation is the variance of the payout:

Var⁡(R)=E[R2]−(E[R])2.\operatorname{Var}(R)=\mathbb{E}[R^2]-\bigl(\mathbb{E}[R]\bigr)^2.

In the example,

E[R2]=0.125(42)+0.375(12)=2.375,\mathbb{E}[R^2]=0.125(4^2)+0.375(1^2)=2.375,

so Var⁡(R)=2.375−0.8752=1.609375\operatorname{Var}(R)=2.375-0.875^2=1.609375 square units. The standard deviation is about 1.271.27 units. Subtracting a fixed one-unit stake does not change the variance, so this is also the variance of the net result.

For mm independent spins with the same rules, expected total net return is mE[X]m\mathbb{E}[X], and variance is mVar⁡(X)m\operatorname{Var}(X). The standard deviation grows as m\sqrt{m}, not as mm. Independence is an assumption here; games with stateful bonuses or changing progressive jackpots may require a more detailed model.

Common modeling mistakes

  • Treating every visible symbol as equally likely. The probabilities may be unequal, and the displayed layout may not specify them.
  • Mixing gross payouts and net winnings. If a payout includes the returned stake, subtract the stake to calculate net profit.
  • Confusing hit rate with profitability. A frequent small prize can still leave the expected net return below zero.
  • Reading expected value as a short-run prediction. It is an average over repeated plays under the model, not a promise about an individual session.

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.