Finding Expected Value- Statistical Definition and Calculation
What Expected Value Actually Is
Expected value (EV) is just the long-term average of a random variable. That's it. If you run an experiment a million times, the expected value is roughly what you'd expect to see per trial.
People overcomplicate this. It's not some mystical prediction—it's a weighted average. You multiply each outcome by its probability, then add everything up.
The Formula
For a discrete random variable:
EV = ÎŁ (x Ă— P(x))
Where x is an outcome and P(x) is its probability.
For continuous random variables, you use an integral instead. Most people dealing with real problems work with discrete cases, so focus there first.
Discrete vs Continuous Expected Value
Discrete Random Variables
These have specific, countable outcomes. Coin flips, dice rolls, stock prices that can only take certain values.
Example: A fair die gives outcomes 1-6, each with probability 1/6.
EV = (1 Ă— 1/6) + (2 Ă— 1/6) + (3 Ă— 1/6) + (4 Ă— 1/6) + (5 Ă— 1/6) + (6 Ă— 1/6) = 3.5
Notice: 3.5 isn't even a possible outcome on a die. That's fine. EV doesn't have to be realistic—it's an average.
Continuous Random Variables
These can take any value in a range. Height, temperature, exact stock prices. You need calculus to work with these properly.
Unless you're doing advanced stats or physics, you won't need this often. Skip it unless your problem specifically requires integration.
Working Examples That Actually Make Sense
Example 1: A Simple Game
You pay $5 to play a game. You flip a fair coin:
- Heads: You win $11
- Tails: You win $0
Is this a good bet?
Winnings: $11 with 0.5 probability, $0 with 0.5 probability
EV of winnings = (11 Ă— 0.5) + (0 Ă— 0.5) = $5.50
You pay $5 to play, so your net EV = $5.50 - $5 = $0.50
On average, you make $0.50 per game. This is a favorable game—barely.
Example 2: Insurance Math
A $500,000 life insurance policy costs $500/year. The probability of a 30-year-old dying in any given year is roughly 0.002 (0.2%).
EV of payout = $500,000 Ă— 0.002 = $1,000
The insurance company pays out $1,000 on average per policy. They charge $500. That's a $500 profit per policy on average.
This is why insurance companies exist. They're not gambling—they're just math businesses.
Example 3: Investment Decision
You have $10,000. Two options:
- Option A: Guaranteed 5% return = $500 profit
- Option B: 70% chance of 20% return, 30% chance of losing everything
Option B EV = (0.70 Ă— $2,000) + (0.30 Ă— -$10,000) = $1,400 - $3,000 = -$1,600
Option B has a negative expected value. You'd lose $1,600 on average. The 70% chance of big gains is a trap—it doesn't offset the catastrophic downside.
How To Calculate Expected Value: Step by Step
Here's the actual process:
- Identify all possible outcomes — List every result that can happen
- Assign probabilities — Determine how likely each outcome is (must sum to 1.0)
- Assign values — What is each outcome worth in your measurement units?
- Multiply and sum — Multiply each outcome by its probability, then add everything
Let's walk through a real scenario: a restaurant deciding whether to add a new menu item.
The item costs $8 to make. They estimate:
- 30% chance it sells well: 100 units/day at $15 profit each = $1,500/day
- 50% chance it sells okay: 40 units/day = $600/day
- 20% chance it flops: 10 units/day = $150/day
EV = (0.30 Ă— $1,500) + (0.50 Ă— $600) + (0.20 Ă— $150) = $450 + $300 + $30 = $780/day
Compare this to the current average day. If the restaurant averages $600/day without the item, adding it gives an expected boost of $180/day. That's worth considering.
Expected Value of a Function
Sometimes you don't care about the raw outcome—you care about a function of it.
E[g(X)] = ÎŁ g(x) Ă— P(x)
Example: You're evaluating a gamble. The outcome X is the dollar amount you win. But you care about utility—your satisfaction with the money. Maybe $0 is devastating but $100 is only mildly good.
You'd apply a utility function g(X) to each outcome before multiplying by probability.
This matters in economics and decision theory. Most basic applications don't need it.
Properties That Actually Help
Linearity
E[aX + b] = aE[X] + b
This works even when X isn't independent. You can break down complex problems into pieces.
Additivity
E[X + Y] = E[X] + E[Y]
Doesn't require independence. This is useful when combining multiple random variables.
Constant Multiplication
E[cX] = cE[X]
Pull constants out of the expectation. Saves calculation time.
Common Mistakes That Ruin Your Calculation
- Forgetting to normalize probabilities — They must sum to 1. If they sum to 0.8, you've missed outcomes or miscalculated
- Confusing EV with most likely outcome — A lottery ticket might have an EV of $0.10, but the most likely outcome is $0
- Ignoring variance — Two investments can have identical EV but wildly different risk profiles
- Using the wrong probability scale — Make sure you're working with decimals (0.5) or percentages (50%), not fractions of fractions
- Not converting to consistent units — If costs are in dollars and benefits in different currencies, convert first
When Expected Value Falls Short
EV is a blunt tool. It ignores:
- Risk tolerance — A 50% chance of losing $1 million might have positive EV, but most people can't afford that risk
- Variance — High variance means results swing wildly around the mean
- Non-linear utilities — Doubling your money doesn't double your happiness
This is why Vegas casinos offer +EV games to customers but customers still lose. The math says you should play—but the variance means you probably won't walk away ahead.
Comparing Decision Frameworks
| Method | What It Considers | Best For |
|---|---|---|
| Expected Value | Average outcome weighted by probability | Repeated decisions, long-term planning |
| Expected Utility | Outcomes adjusted for risk preference | Decisions with significant downside risk |
| Worst Case / Best Case | Only extremes | Risk-averse decisions, one-time bets |
| Median Outcome | Middle probability point | When average misleads due to outliers |
Where You'll Actually Use This
Finance: Pricing options, evaluating portfolios, insurance underwriting
Gambling: Determining if a bet has positive EV (spoiler: most don't)
Business: Pricing strategies, investment decisions, inventory management
Science: Estimating population parameters, hypothesis testing foundations
Everyday decisions: Whether to upgrade your phone plan, take a job offer with commission vs salary, buy extended warranties
The Bottom Line
Expected value is just weighted average. You calculate it by multiplying each outcome by its probability and adding the results.
It doesn't tell you what will happen—it tells you what to expect on average over many repetitions. That's useful information, but it's not a prediction.
Use it when outcomes are repeatable and you care about long-term averages. Question it when outcomes are one-time, variance is high, or risk matters more than the number.