An Average Is Not a Single Result
In Choosing Between Options Using Expected Value, you used expected value to compare the modeled average costs of different options. That average is useful for decisions, but it is not a prediction of exactly what will happen on one play. A game can have a positive expected gain even when a player is more likely to lose than to win.
The expected value \(\mu_X\) of a random variable \(X\) is a probability-weighted average of its possible outcomes. As explained in Interpreting Expected Value as a Long-Run Average, it describes the average outcome over many repetitions of the same chance process under comparable conditions. It does not have to be one of the values \(X\) can take, and it does not tell us which value will occur on the next play.
For a one-play decision, risk includes what outcomes are possible, how likely they are, and how large a gain or loss could be. Two games can have the same expected gain but very different single-play risks. A player might receive a small gain for certain in one game, while another game offers a chance of a larger gain and a chance of a loss.
Expected value alone does not say which option every person should choose. It provides a comparison of average outcomes under the model. Someone deciding whether to play may also care about the chance of losing, the size of a possible loss, and whether they can afford that loss. Keep those questions separate: “Which option has the higher expected gain?” is not the same as “Which option is less risky for me on one play?”
What the Long-Run Average Does—and Does Not—Say
Imagine repeating the same game many times, with each play following the same probability model. The average net gain per play across those repetitions tends to be close to the expected gain as the number of plays grows. This long-run interpretation does not promise that the average will equal the expected value after a particular number of plays. Nor does it guarantee that a player will have a positive total gain.
Short runs can be uneven. A player may get an unusual sequence of outcomes, especially when an outcome is rare. Even after many plays, the actual average may differ from the expected value. The expected value is the center specified by the model, not a schedule that forces results to balance out after a fixed number of plays.
It is also important to distinguish the average outcome from the most likely outcome. A low-probability, large gain can pull an expected value upward even if most plays lose money. The distribution contains the full story: its possible values and their probabilities show both the average and what can happen on one play.
A Practical Way to Discuss Expected Value and Risk
Specify whether the random variable represents a prize, a cost, or net gain per play. Include the entry fee when calculating net gain.
Identify every possible result and its probability. Check that the probabilities are between 0 and 1 and add to 1, as in Checking Whether a Probability Distribution Is Valid.
Use the probability-weighted average from The Mean of a Discrete Random Variable. State the average in context and with units.
State the possible outcomes, including losses, and their probabilities. Explain which outcome occurs on one play is uncertain.
Use the chances and sizes of gains and losses to describe what a player might experience. Do not claim that the expected value guarantees the result of one play.
The word average needs a clear reference. If \(X\) is net gain in dollars on one play, \(\mu_X\) is the expected net gain per play. It is not the total amount a player will earn, and it is not necessarily an amount that can occur on any play. Keep the units and time period attached to the interpretation.
Worked Example: A Positive Average with a Likely Loss
Worked Example: A Positive Average with a Likely Loss
A fictional game costs \(\$3\) to play. A player receives a \(\$15\) prize with probability \(0.25\), and receives no prize with probability \(0.75\). Compare the expected net gain with the result of one play.
State. Let \(X\) be the player’s net gain, in dollars, from one play. We want to describe both the average net gain in the model and the possible outcome on a single play.
Plan. Net gain is prize minus the \(\$3\) entry fee. The possible net gains are \(\$12\) when the prize is won and \(-\$3\) when there is no prize. The probabilities \(0.25\) and \(0.75\) are between 0 and 1 and sum to 1; these two outcomes cover all possibilities in the stated model. We will calculate the probability-weighted average, then compare it with the possible one-play results.
Do. The expected net gain is:
The calculation checks by separating the expected prize from the entry fee: the expected prize is \((15)(0.25)+(0)(0.75)=\$3.75\), and \(\$3.75-\$3.00=\$0.75\). On one play, however, the net gain is either \(\$12\) or \(-\$3\), not \(\$0.75\). The player has a \(0.75\) probability of losing \(\$3\) on that play.
Conclude. Under this model, the expected net gain is \(\$0.75\) per play over many comparable repetitions. That positive average does not mean the player will gain \(\$0.75\) on one play; the player is more likely to lose \(\$3\) than to win \(\$12\) on a single play.
Worked Example: A Small Expected Gain and a Frequent Loss
Worked Example: A Small Expected Gain and a Frequent Loss
A fictional arcade challenge costs \(\$1\). On a play, the player either earns a \(\$0.50\) prize with probability \(0.20\) or earns no prize with probability \(0.80\). Find and interpret the expected net gain, and describe the risk on one play.
Define the variable. Let \(X\) be the player’s net gain, in dollars, on one play. Winning the prize gives a net gain of \(\$0.50-\$1.00=-\$0.50\). Receiving no prize gives a net gain of \(\$0-\$1.00=-\$1.00\). The probabilities \(0.20\) and \(0.80\) are valid and total 1, and the two outcomes account for every play in this model.
Calculate. Weight each net gain by its probability and add:
A second calculation uses expected prize minus the entry fee. The expected prize is \((0.50)(0.20)+(0)(0.80)=\$0.10\), so the expected net gain is \(\$0.10-\$1.00=-\$0.90\). Both calculations agree.
Interpret the result. The expected net gain is a loss of \(\$0.90\) per play on average over many repetitions under this model. On one play, the player loses either \(\$0.50\) or \(\$1.00\); even winning the prize does not cover the entry fee. The chance of losing \(\$0.50\) is \(0.20\), and the chance of losing \(\$1.00\) is \(0.80\). Here, the expected value summarizes the average loss, while the distribution shows the different losses possible on an individual play.
Worked Example: Same Expected Gain, Different Risk
Worked Example: Same Expected Gain, Different Risk
A student has two fictional choices for a one-play challenge. Choice A gives a net gain of \(\$1\) for certain. Choice B gives a net gain of \(-\$4\) with probability \(0.50\) and \(\$6\) with probability \(0.50\). Compare expected gains and describe the difference in risk.
Check the models. Choice A has one possible outcome, \(\$1\), with probability 1. Choice B’s two probabilities are between 0 and 1 and sum to 1, and its two outcomes cover the model’s possibilities.
Calculate the expected gains. For Choice A:
For Choice B:
Both choices have an expected gain of \(\$1\) per play. Yet the outcomes differ: Choice A always gives a \(\$1\) gain, while Choice B has a \(0.50\) probability of a \(\$4\) loss and a \(0.50\) probability of a \(\$6\) gain.
The standard deviations, from the method in The Standard Deviation of a Discrete Random Variable, also describe a difference in spread. Choice A has standard deviation \(\$0\), because it always gives its mean. For Choice B, the mean is \(\$1\), so each outcome is \(\$5\) from the mean. Thus:
Conclude. The two choices have the same expected gain, but not the same one-play risk. Choice A guarantees a \(\$1\) gain under its model; Choice B can produce either a \(\$4\) loss or a \(\$6\) gain. The expected-value comparison alone does not determine which choice is preferable to a particular student.
Common Mistakes and AP Exam Tips
- Treating expected value as a promised result. A model’s expected gain is a probability-weighted average, not a guaranteed gain on the next play. A full-credit answer distinguishes the long-run average from the possible individual outcomes.
- Ignoring the entry fee when discussing gain. A prize is not the same as net gain. Subtract the cost to play from the prize for each outcome before interpreting expected net gain.
- Assuming a positive expected value means most plays win. A sufficiently large, uncommon gain can outweigh frequent losses in the weighted average. Read the probabilities and outcome sizes, not just the mean.
- Calling equal expected values equal risk. Equal means do not imply equal distributions. Name the possible outcomes and their chances; if useful, compare standard deviations as a measure of spread.
- Describing risk only with a mean. A mean does not identify the chance or size of a loss. A clear risk description states the loss amount and its probability in context.
- Claiming that results must balance after a set number of plays. The expected value is a long-run average interpretation, not a guarantee about any fixed number of repetitions. Actual short-run results can differ from the model’s average.
A strong AP response makes the variable and units clear, interprets the expected value as an average over many comparable plays, and then describes the one-play possibilities. For example: “The expected net gain is \(\$1\) per play, but this is a long-run average under the model. On one play, Choice B has a \(0.50\) chance of losing \(\$4\) and a \(0.50\) chance of gaining \(\$6\).”
Key Takeaway
Expected value summarizes a distribution’s probability-weighted average, while a single play produces one particular outcome. Use the distribution to discuss both: report the average in context, and describe the chances and sizes of possible individual gains and losses.
Check Your Understanding
Use the fictional game below. It costs \(\$2\) to play and awards a \(\$10\) prize with probability \(0.10\); otherwise, there is no prize.
- Define \(X\) as net gain on one play and state its possible values and probabilities.
- Check that the probabilities form a valid distribution, then calculate \(\mu_X\).
- Interpret the expected net gain in context, including the relevant time frame.
- What is the probability of losing money on one play, and how much is lost in that outcome?
- Explain why the expected net gain does not predict the result of the next play.