What is expected value?

Expected value, usually shortened to EV, is the average amount a bet or decision is worth after every possible outcome is weighted by its probability. It does not predict the next hand, spin, roll, or session. It answers a different question: if the same decision could be repeated under the same conditions many times, what would the average result be per decision?

That distinction matters because casino outcomes are noisy. A negative-EV bet can win immediately. A correct blackjack decision can lose. One result tells you what happened once; expected value tells you how the price of the decision behaves across repetition.

The formula behind the plain-English definition

For outcomes numbered from 1 to n:

$$EV = \sum_{i=1}^{n} p_i x_i$$

Where:

• $p_i$ is the probability of outcome i;
• $x_i$ is the net financial result of that outcome;
• the probabilities must add to 1;
• wins are positive, losses are negative, and pushes are zero.

The word net prevents a common error. A $10 wager that returns $360 after a straight-up roulette win produced a $350 profit, not a $360 profit, because $10 of the return is the original stake.

A complete roulette example

Consider a $10 straight-up wager on one number on an American double-zero wheel. There are 38 pockets. The bet wins on one pocket and loses on the other 37. The posted payout is 35 to 1.

So:

$$EV = \left(\frac{1}{38} \times 350\right) + \left(\frac{37}{38} \times -10\right) = -0.5263$$

The expected value is approximately -$0.53 per $10 wager. Dividing that expected loss by the initial $10 stake gives the familiar 5.26% house edge:

$$House\ Edge = \frac{0.5263}{10} = 5.263%$$

The number can still hit on the first spin. If it does, the player wins $350. That result does not change the calculation because the calculation already includes the possibility of winning.

EV is not probability, payout, or house edge

These terms are related but not interchangeable:

• Probability describes how often an outcome is expected to occur.
• Payout describes what the bet returns when it wins.
• Expected value combines every probability with every net result.
• House edge expresses the casino’s expected advantage as a percentage of the initial wager.
• Expected loss applies the edge to total action over many wagers.
• Variance describes how widely actual results can move around the expected value.

A large payout can coexist with poor EV when the winning probability is too small. A high hit frequency can coexist with poor EV when many wins return less than the amount risked. EV is useful precisely because it refuses to judge either probability or payout in isolation.

Fair payout is an EV question

Expected value also shows whether a payout is fair. If a wager wins with probability $p$ and otherwise loses one unit, the fair net payout $r$ satisfies:

$$p imes r-(1-p)=0$$

So:

$$r=\frac{1-p}{p}$$

For a one-number wager on a 38-pocket wheel, $p=1/38$. A fair net payout would therefore be 37 to 1. The actual payout is 35 to 1. The missing two units across the 38 equally likely pockets create the casino advantage.

This method is valuable when a paytable looks impressive. Instead of asking whether the top prize is large, compare the posted payout with the probability required to earn it. A bet can have several winning outcomes, pushes, bonus returns, and partial losses; in that case, all branches must be included rather than forcing the wager into a simple win-or-lose formula.

From one wager to a session

Expected value is additive. If the conditions and stake remain the same, repeating a -$0.5263 wager 100 times produces an expected total result of:

$$100 \times -0.5263 = -52.63$$

The same answer can be reached through total action:

$$Total\ Action = 100 \times 10 = 1{,}000$$

$$Expected\ Loss = 1{,}000 \times 0.05263 = 52.63$$

This does not mean every 100-spin session loses $52.63. One session may finish $300 ahead; another may lose the full $1,000. EV gives the center of the long-run distribution. Variance explains why actual sessions can land far from that center.

The practical implication is that speed matters. A player making twice as many identical negative-EV wagers creates roughly twice the expected loss, even though the house-edge percentage has not changed. That is why total action and expected loss in real sessions are more useful than staring only at the edge percentage.

Decision quality and outcome quality are different

Suppose a blackjack hand should be doubled under the table’s rules and basic strategy. The player doubles and loses. The outcome is bad, but the decision can still have higher EV than hitting once or standing.

The reverse also happens. A player makes a poor side bet and wins. The outcome is good, but the decision can still have strongly negative EV.

This is why statements such as “I won, so the bet was smart” and “I lost, so the decision was wrong” are unreliable. Results are evidence only when the sample is large enough and the conditions are controlled. A single outcome is not an audit of the underlying mathematics.

Can casino EV ever be positive for the player?

Normal casino wagers are designed to have negative player EV. Positive EV can arise only when something changes the total value of the decision, such as:

• a promotion or rebate large enough to exceed the base disadvantage;
• a progressive jackpot that has grown beyond a mathematical threshold;
• a genuine skill or information advantage;
• a rules or payout error;
• a tournament structure in which prizes exceed total entry costs.

The conditions must be counted honestly. Travel cost, eligibility restrictions, wagering requirements, taxes, limited availability, variance, and the risk of making strategy errors can turn an apparent advantage into a negative one. “There is a promotion” is not the same as “the complete decision is +EV.”

Why casino operators use EV

A casino does not expect every table, machine, or player to finish near theoretical result each night. Operations use expected value as a baseline for planning and review. A table-game rating can estimate theoretical loss from average bet, decisions per hour, time played, and the game edge. A slot system can use coin-in and theoretical hold. Finance and marketing can compare actual results with theoretical results over suitable periods.

That comparison must be handled carefully. A large actual win or loss can be normal variance. A persistent gap can also point to rating errors, downtime, incorrect game configuration, settlement mistakes, unrecorded promotions, or an unsuitable model. EV is therefore a benchmark for investigation, not an accusation that every short-term result should match the model.

What EV cannot tell you

Expected value does not tell you:

• what will happen next;
• how large your worst short-term loss can be;
• how much bankroll is needed to tolerate volatility;
• whether you will reach the long-run average during your lifetime;
• whether a gambling expense is affordable;
• whether continuing after a loss is sensible.

Those questions require variance, bankroll limits, time horizon, and personal affordability. EV is a decision-pricing tool, not a promise and not a safety guarantee.

Common EV calculation mistakes

Several errors can reverse or distort the answer:

• using the total return instead of net profit on a winning outcome;
• omitting pushes, partial returns, fees, commissions, or bonus conditions;
• using probabilities that do not add to 1;
• mixing per-hand EV with per-dollar or per-session EV;
• applying a house edge from one rule set to another;
• treating a theoretical average as a short-term prediction;
• comparing two bets with different amounts at risk without stating the denominator.

A sound calculation should be reproducible. Another reader should be able to see the outcomes, probabilities, net results, and arithmetic—not just a final percentage. For a formal statistics treatment, OpenStax explains expected value as the probability-weighted long-run mean of a discrete random variable.

A practical way to use expected value

Before comparing two wagers, make the comparison on the same basis:

1. Confirm the exact rules and paytable.
2. List all outcomes, including pushes and partial returns.
3. Use net wins and net losses.
4. Multiply each result by its probability.
5. Add the weighted results.
6. State the denominator: per initial wager, per total amount risked, per hand, or per session.
7. Keep EV separate from variance and short-term luck.

For readers who want the compact mathematical definition, use the Expected Value glossary entry. To convert an edge into a realistic session estimate, use the expected loss calculator and read Why Session Luck Hides Long-Term Math.

The essential point is simple: expected value evaluates the price of a repeatable decision before the result is known. A lucky result cannot repair bad pricing, and an unlucky result cannot by itself disprove sound mathematics.