Mathematics & Logic · Probability & statistics
What an expected value promises
An expected value is the average you'd get over many repetitions. See how to calculate one, why it can be a result that never happens, and what it can't tell you about a single try.

What you will learn
- State what an expected value claims and what it doesn't
- Describe the law of large numbers and what it does and doesn't guarantee
- Build an expected value table and compute the mean of a discrete distribution
- Explain why an expected value doesn't have to be a possible outcome
Lesson 1 of 4 · 5 min
A long-run average, not a forecast
Toss a fair coin ten times and you might get nine heads. Nothing has gone wrong with the coin. Probability tells you what to expect over the long run, and very little about a short one.
The same goes for an expected value. It's the average result you'd get from repeating something a great many times, which is why it's also called the long-term average. Statisticians write it with the Greek letter mu.
So an expected value is never a prediction of the next outcome. A run of losses doesn't prove the calculation wrong. A run of wins doesn't prove the game is in your favor.
What settles down with repetition is the relative frequency, the share of trials in which an event happens. The law of large numbers says that as trials pile up, the gap between that share and the true probability shrinks toward zero. Averages behave the same way and close in on the expected value.
Karl Pearson once tossed a coin 24,000 times and recorded every toss. He got 12,012 heads. That's a share of 0.5005, against a probability of 0.5.
Look at what converged. The share of heads landed within a hair of one half, but the count didn't even out. Pearson finished with 24 more heads than tails, and nothing pulls that surplus back to zero. In a longer run the gap in counts tends to grow, even as the share closes in on a half. Heads and tails are never due.
Try to recallState the law of large numbers, and say whether it promises that the number of heads will approach the number of tails.
The law of large numbers says that as the number of trials increases, the gap between an event's relative frequency and its probability approaches zero. It's a statement about the share, not the counts. Pearson tossed a coin 24,000 times and got 12,012 heads, a share of about 0.5005, while still finishing with 24 more heads than tails. Nothing pulls that difference back toward zero, so the law doesn't promise that heads and tails will even out in number.
An expected value is a promise about the average of many repetitions. Pearson needed 24,000 tosses to land within 0.0005 of one half. It promises nothing about the next toss, or the next ten.
Lessons in this topic
A long-run average, not a forecast
An expected value says what happens over many repetitions and nothing about the next one. A coin tossed 24,000 times shows the difference.
Calculating an average that never actually happens
Multiply each outcome by its probability and add. The result is often a number the experiment can never produce, like 3.5 on a die.
The average loss that no player ever pays
Work through a 2-dollar game with a 100,000-dollar prize. Its expected result is a loss of 1 dollar, which never happens to anyone who plays.
Standard deviation, the spread an average hides
Add one column to the expected value table to get a standard deviation, and see how two games with the same average can be nothing alike.
Three quick stories from this topic
Pearson's 24,000 coin tosses
Karl Pearson tossed a coin 24,000 times and got 12,012 heads
That's a share of 0.5005, within a hair of one half. Yet he finished with 24 more heads than tails. The law of large numbers pulls the share of heads toward a half, and it does nothing to even out the count.
What the law says
As the number of trials increases, the difference between the theoretical probability and the relative frequency approaches zero. That's the law of large numbers.
What it doesn't say
The proportion converges, but the counts don't. Pearson was still twelve heads clear of an even split, and nothing pushes that surplus back toward zero.
The loss nobody pays
A lottery can cost players 1 dollar a play on average without anyone ever losing 1 dollar
In a textbook game you pay 2 dollars and gain 100,000 if five digits match, a 1 in 100,000 chance. The average result is a loss of about 1 dollar. But every play ends with you 2 dollars down or 100,000 up.
The catch
Each time you play, you either lose 2 dollars or profit 100,000. No play of this game ever costs you 1 dollar. The 1 dollar is the average after playing over and over.
The question to ask
Ask how many repetitions your decision actually involves. An expected value is a promise about the long run and nothing else.
A die's expected roll is 3.5
The expected value of a die roll is 3.5, a number no die can show
Add the faces from 1 to 6 and divide by 6, since each is equally likely. Roll many times and your average closes in on that figure. An expected value is a long-run average, so it doesn't have to be a result you can get.
It's a weighted average
Multiply each outcome by its probability and add the products. The total lands wherever the weights balance, which is often between the outcomes.
It isn't the typical case
A team that plays 0, 1 or 2 days a week can have an expected 1.1 days. No week looks like that, and the calculation is still right.
Sources
- 4.2 Mean or Expected Value and Standard Deviation, OpenStax, Introductory Statistics 2e
- Measures of the Center of the Data, OpenStax, Rice University
- Law of large numbers, Wikipedia
- Expected value, Wikipedia





