Opening subject page...
Loading your content
Learn how to use probability and expected value to make smarter decisions when outcomes are uncertain.
Long before statisticians formalized the concept of expected value, people were already grappling with uncertainty. Merchants weighed the risk of shipping goods across dangerous seas, gamblers argued over fair divisions of stakes in interrupted games, and generals calculated whether a bold strategy was worth the potential losses. The mathematical tools we use today to evaluate decisions under uncertainty grew directly from these real-world dilemmas.
The central question has always been the same: when you face a choice and the outcomes depend on chance, how do you figure out which option is best in the long run? Expected value provides a principled, quantitative answer — and its history reveals how some of the greatest mathematical minds tackled this very problem.
The thread connecting all of these milestones is a single, powerful insight: when you can't know the future for certain, you can still make rational choices by computing the long-run average outcome of each option. That computation is what expected value is all about, and learning to use it will sharpen every decision you make under uncertainty.
Before you can evaluate and compare strategies, you need a rock-solid understanding of several foundational ideas. Each one builds on the last, so take them in order.
The following diagram shows how expected value is calculated for a simple scenario: a carnival game where you spin a wheel with four possible payouts. Each sector of the wheel has a different size (representing probability) and a different dollar value. Expected value is found by multiplying each payout by its probability and summing the results.
Notice a few things about this diagram. First, the $0 outcome has the largest sector — 40% of the wheel — but it contributes nothing to the expected value because $0 × 0.40 = $0. Second, the $25 outcome is exciting when it hits, but it only makes up 10% of the wheel, so it contributes just $2.50. The expected value of $5.50 is not one of the actual outcomes you can land on — it's the theoretical long-run average. If the carnival charges $6 to spin, the game is a losing proposition on average because your expected payout ($5.50) is less than the cost ($6). That difference — the gap between what you pay and what you expect to win — is how carnivals (and casinos, and insurance companies) make their money.
Now that you have the intuition, let's formalize the math. The expected value of a discrete random variable X is calculated with the following formula.
In plain language: for every possible outcome, multiply the value of that outcome by the probability it occurs, then add all those products together. The result is the expected value, sometimes also written as μ (the Greek letter mu). Let's break the formula into its components so you know exactly what each piece means.
xi represents each distinct value the random variable can take. In a coin-flip bet where you win $10 for heads and lose $5 for tails, the possible values are x₁ = +10 and x₂ = −5. P(xi) is the probability of that specific outcome. For a fair coin, P(+10) = 0.5 and P(−5) = 0.5. The Σ (sigma) symbol means "sum up all of these products."
This comparison rule is the heart of decision-making with expected values. Suppose you're choosing between two insurance plans, two investment options, or two game strategies. You compute E(A) and E(B) separately, then compare them. The strategy that yields a higher expected gain (or lower expected cost) is the better choice on average over many repetitions.
There are two important properties of expected value that simplify calculations. First, expected value is linear: E(aX + b) = a × E(X) + b, where a and b are constants. This means if every payout in a game doubles, the expected value doubles too. Second, for independent events, E(X + Y) = E(X) + E(Y) — you can calculate expected values separately and add them. These properties make it possible to analyze complex, multi-stage strategies by breaking them into simpler parts.
When real decisions involve multiple stages or branching possibilities, a decision tree is the standard tool for organizing the calculation. A decision tree lays out each choice you can make, the chance events that follow, and the payoff at each endpoint. You then work backward from the endpoints, computing expected values at each branch, until you arrive at the expected value of each initial strategy.
The following diagram shows a decision tree for a student deciding between two summer job strategies. Strategy A is a guaranteed hourly job paying a fixed amount. Strategy B is a commission-based sales internship where earnings depend on performance — which is uncertain.
The decision tree reveals that Strategy A (the hourly job) has an expected value of $3,200, while Strategy B (the sales internship) has an expected value of $3,150. By the expected value criterion, Strategy A is the slightly better choice. However — and this is important — notice that Strategy B offers a 30% chance of earning $5,000, which is far more than Strategy A can deliver. If you can afford the risk and the upside matters to you, a reasonable person might still pick B. Expected value tells you the long-run average, but it doesn't capture everything about a decision, especially when you only get to make the choice once.
Let's walk through a complete, realistic problem. Suppose you're helping a family choose between two health insurance plans. Plan X has a lower monthly premium but higher out-of-pocket costs when health events occur. Plan Y has a higher premium but covers more. Given the family's health history, their doctor estimates the following probability distribution for annual medical expenses (beyond what insurance covers).
| Scenario | Probability | Plan X Out-of-Pocket | Plan Y Out-of-Pocket |
|---|---|---|---|
| Healthy year (no major events) | 0.50 | $500 | $200 |
| Moderate health events | 0.35 | $3,000 | $800 |
| Major health event | 0.15 | $8,000 | $1,500 |
Expected value is a powerful criterion for comparing strategies, but like any tool, it has situations where it shines and situations where it falls short. Understanding both sides will make you a more thoughtful decision-maker.
| Strengths | Limitations |
|---|---|
| Provides a single, objective number for comparison — eliminates guesswork and emotional bias. | Ignores variability (spread) of outcomes. Two strategies can have the same EV but wildly different risk profiles. |
| Guaranteed to reflect long-run averages (Law of Large Numbers). Ideal for repeated decisions (e.g., a business making the same choice hundreds of times). | Less reliable for one-time decisions where you can't "average out" bad luck. The long run may never arrive. |
| Mathematically clean — linear, additive, easy to compute. Works with decision trees and complex multi-stage problems. | Does not account for diminishing marginal utility. Gaining $1,000,000 doesn't feel twice as good as gaining $500,000, yet EV treats it that way. |
| Universal applicability: finance, medicine, sports, engineering, everyday life. | Requires accurate probability estimates. If your probabilities are wrong, your expected value is wrong too. |
Expected value is the starting point of a much richer world of decision theory. As you advance in statistics and probability, you'll encounter several extensions that address the limitations we discussed.
| Concept | What Expected Value Does | What the Advanced Version Adds |
|---|---|---|
| Expected Utility Theory | Treats every dollar as equally valuable — $1,000 always matters the same amount regardless of your wealth. | Replaces dollar values with utility values that account for diminishing returns. A risk-averse person values the certainty of $500 more than a 50/50 chance at $1,000, even though the EV is the same. |
| Variance & Standard Deviation | Gives only the center (mean) of the outcome distribution. | Measures how spread out outcomes are around the expected value. Two strategies with the same EV but different variances carry different levels of risk. |
| Conditional Expected Value | Uses fixed, unconditional probabilities. | Updates probabilities when new information arrives (using Bayes' Theorem), producing expected values that adapt to what you've learned. |
| Game Theory | Evaluates your strategy in isolation. | Considers how other players' strategies affect your outcomes. Nash equilibrium finds optimal strategies when everyone is maximizing their own expected value simultaneously. |
For now, the core expected value framework gives you everything you need to evaluate and compare strategies under uncertainty. As you encounter more complex situations — in AP Statistics, college-level courses, or real life — these advanced tools will build naturally on the foundation you're developing here. The logic is always the same: assign probabilities to outcomes, compute a weighted average, and use it to make informed decisions.
Test your understanding with these five problems, ordered from conceptual to challenging. Try each one before revealing the answer.
In this lesson, you learned that expected value is the probability-weighted average of all possible outcomes of a random event — computed by the formula E(X) = Σ [xᵢ × P(xᵢ)]. This single number captures the long-run average result of a strategy, making it the standard tool for comparing options under uncertainty. You saw how decision trees organize complex, multi-stage decisions by mapping out choices, chance events, and payoffs, then working backward to compute expected values at each branch. Through worked examples — from carnival wheels and insurance plans to farming decisions and basketball strategy — you practiced calculating expected values and using them to determine which strategy is objectively better in the long run.
You also learned that expected value has important limitations: it doesn't account for how spread out (risky) the outcomes are, and it assumes every dollar is equally important regardless of context. For one-time, high-stakes decisions, risk tolerance and worst-case analysis should supplement expected value. Advanced extensions like expected utility theory and variance analysis address these gaps, building on the foundation you now have. The core skill — translating uncertain situations into probability distributions, computing weighted averages, and comparing strategies quantitatively — is one of the most practical tools in all of statistics.