How to Calculate Expected Values from Data

What Is Expected Value and Why Should You Care?

Expected value (EV) is the long-run average outcome you'd expect if you repeated an experiment infinite times. It's not a prediction of what will happen in a single trial. It's a mathematical center of gravity for your data.

If you're working with probability distributions, running simulations, or making decisions under uncertainty, you need expected values. Period.

The Basic Formula

For a discrete random variable X with possible values x₁, x₂, x₃... and their probabilities p₁, p₂, p₃...

EV = Σ(xᵢ × pᵢ)

That's it. Multiply each outcome by its probability, then add everything up.

For continuous random variables, you swap the sum for an integral:

EV = ∫x × f(x) dx

Where f(x) is your probability density function.

Calculating Expected Value from Raw Data

Most of the time you don't have a nice formula. You have a dataset. Here's how to handle it:

Method 1: Grouped Data with Frequencies

When you have frequency counts, treat frequency as probability weight:

EV = Σ(xᵢ × fᵢ) / n

Where fᵢ is the frequency and n is total observations.

Method 2: Direct Sample Mean

When every observation is equally likely (which is usually the case in raw data), the sample mean is your expected value:

EV ≈ x̄ = (1/n) × Σxᵢ

This works because each data point represents one outcome from your distribution. The mean captures the expected value by definition.

Method 3: From a Probability Distribution

If you've fit a distribution to your data (normal, Poisson, binomial), plug the parameters into the known EV formula for that distribution.

Step-by-Step: Calculating EV from a Dataset

Let's work through a real example. Say you're analyzing daily sales numbers:

Data: 45, 52, 48, 61, 39, 55, 50, 47, 53, 49

Step 1: Count your observations (n = 10)

Step 2: Sum all values (45+52+48+61+39+55+50+47+53+49 = 499)

Step 3: Divide by n (499 / 10 = 49.9)

Your expected daily sales value is 49.9 units.

That's the straightforward method. But what if your data has different frequencies?

Weighted Data Example

Say you're analyzing customer purchase amounts and you have grouped data:

• $0-50: 120 customers
• $51-100: 85 customers
• $101-150: 45 customers
• $151-200: 20 customers

Use the midpoint of each range as your value:

EV = (25 × 120 + 75 × 85 + 125 × 45 + 175 × 20) / 270

EV = (3000 + 6375 + 5625 + 3500) / 270

EV = 18500 / 270 = $68.52

Expected Value of a Function of Random Variables

Sometimes you need E[f(X)], not E[X]. The good news: there's a direct formula.

E[f(X)] = Σ f(xᵢ) × pᵢ

Example: You want expected profit, not expected sales. Profit = Sales × 0.3 (30% margin).

E[Profit] = 0.3 × E[Sales] = 0.3 × 49.9 = $14.97

You don't need to recalculate from scratch. Just multiply the EV by the constant.

Common Mistakes That Kill Your Calculations

• Forgetting to normalize: Frequencies must become probabilities (divide by total). Raw frequencies give wrong results.
• Confusing population vs sample: The sample mean estimates population EV. It's not the true EV unless n is your entire population.
• Using median when you need mean: They're not the same. Median is the 50th percentile. EV is the probability-weighted average.
• Ignoring outliers: Extreme values heavily influence EV. A single $1M transaction skews your expected revenue hard.
• Misidentifying the random variable: Make sure you're calculating EV of the right thing. E[demand] vs E[revenue] are different calculations.

When Expected Value Misleads You

EV is useless in some situations. If you're evaluating a lottery ticket:

• 99% chance: lose $10
• 1% chance: win $500

EV = (0.99 × -10) + (0.01 × 500) = -9.9 + 5 = -$4.90

The EV is negative. But that doesn't tell you what to do. It tells you the average outcome over many repeated plays. A single ticket is a one-shot decision. EV doesn't capture your risk tolerance or budget constraints.

Use EV for repeated decisions. Use utility functions or risk-adjusted metrics for one-off high-stakes choices.

Quick Reference: Python Implementation

import numpy as np

# Method 1: Sample mean (raw data)
data = [45, 52, 48, 61, 39, 55, 50, 47, 53, 49]
ev = np.mean(data)
print(f"Sample EV: {ev}")

# Method 2: Weighted calculation
values = [25, 75, 125, 175]
weights = [120, 85, 45, 20]
ev_weighted = np.average(values, weights=weights)
print(f"Weighted EV: {ev_weighted}")

# Method 3: From probability distribution
# If your data follows a Poisson distribution with lambda=50
from scipy import stats
ev_poisson = 50  # Lambda IS the EV for Poisson

Bottom Line

Expected value is a weighted average. That's all it is. Calculate it from raw data by taking the mean. Calculate it from a distribution using the appropriate formula. Don't overthink it.

Just make sure you're computing EV for the right variable, with proper probability weights, and that EV is actually the right metric for your decision context. It often isn't.