Expected value is one of the most misunderstood ideas in probability. Students often hear the word “expected” and think it means what will probably happen. But expected value is not the most likely outcome. It is a weighted average over all possible outcomes.
That distinction matters.
If a lottery ticket has a one-in-a-million chance of paying a million dollars and otherwise pays nothing, the expected value may be close to one dollar. That does not mean you should expect to walk away with one dollar. You will almost certainly get zero. The expected value describes the long-run average if the same situation were repeated many times.
A simple game
Suppose you roll a fair six-sided die. If it lands on 6, you win $12. Otherwise, you win nothing.
The expected value is
(1/6)(12) + (5/6)(0) = 2.
So the game has expected value $2.
This means that over many plays, the average payout per play approaches 2. In fact, no single play pays $2.
That is the first important lesson: expected value can be a number that is not itself an outcome.
Why weights matter
Expected value multiplies each outcome by its probability:
E = p1*x1 + p2*x2 + ... + pk*xk.
Outcomes with higher probability matter more. Outcomes with larger values matter more. Expected value balances both.
This is why a rare huge payoff can affect the average even if it is unlikely. It is also why a frequent small loss can dominate a rare win.
Fair games
A game is fair if the expected gain is zero. Suppose a carnival game costs 12 only on a roll of 6.
The expected payout is 3. So the expected gain is
2 - 3 = -1.
On average, the player loses $1 per play.
This does not mean every player loses. Some win. But the structure favors the game operator over repeated play.
This is a useful applied lesson because expected value lets students analyze games, insurance, pricing, and risk.
Linearity of expectation
The most powerful fact about expected value is linearity:
E(X + Y) = E(X) + E(Y).
This works even when X and Y are not independent.
That surprises students because many probability rules require independence. Expected value is different. The average of a sum is the sum of the averages.
Counting fixed points
Here is a classic example.
Five students randomly receive five papers, one paper each. What is the expected number of students who receive their own paper?
Instead of listing all permutations, define indicator variables.
Let Xi = 1 if student i gets their own paper, and 0 otherwise.
The number of students who get their own paper is
X1 + X2 + X3 + X4 + X5.
Each student has probability 1/5 of getting their own paper, so
E(Xi) = 1/5.
By linearity,
E(X1 + X2 + X3 + X4 + X5) = 5*(1/5) = 1.
The expected number of fixed points is 1.
This is elegant because the events are not independent, but expectation still handles them cleanly.
Why indicators help
An indicator variable turns a yes-or-no event into a number. It equals 1 if the event happens and 0 if it does not.
The expected value of an indicator is the probability of the event:
E(I) = 1*p + 0*(1-p) = p.
This lets us count complicated things by adding simple yes-or-no questions.
Instead of counting the whole object directly, we ask: for each possible feature, what is the chance it appears?
A COSMO activity
Ask students:
A coin is flipped 10 times. What is the expected number of heads?
Most students can answer 5. Then ask them to justify it with indicators.
Let Xi = 1 if flip i is heads. Then each E(Xi)=1/2, and the total number of heads is the sum of the indicators. So the expected total is 10/2 = 5.
Now change the problem:
A coin is flipped 10 times. What is the expected number of adjacent HH pairs?
Define Yi = 1 if flips i and i+1 are both heads. There are 9 adjacent pairs. Each has probability 1/4. So the expected number is
9/4.
This example feels harder, but linearity makes it manageable.
Expected value is not a guarantee
Expected value should not be confused with safety. A game can have positive expected value and still carry high risk. A strategy can be good on average and terrible in worst cases.
This is where expected value connects to optimization. Choosing by expected value means choosing by average outcome. Sometimes that is appropriate. Sometimes the spread, risk, or worst-case outcome matters more.
Students should learn to ask what average reasoning ignores.
What to remember
Expected value is a weighted average, not a prediction of the most likely single outcome. Its most useful property is linearity, which lets us break complicated counts into simpler pieces.
The strongest COSMO takeaway is this:
Probability is not only about calculating chances after a situation is described. It can help us design variables that make the structure of the situation easier to see.