Expected Value

Expected value is the theoretical long‑run average of a risky choice, found by weighting each possible result by its chance. For example, a 1% chance to win $100 has an expected value of about $1, so paying $2 for that ticket is unfavorable. Note: In few repetitions, actual results may differ significantly from this average. Try this: Pick a daily choice, list two plausible outcomes with rough odds, multiply and add to get the average.

Expected value compresses uncertain outcomes into a theoretical long‑run average to compare options on equal footing. It provides a consistent yardstick across choices.

You estimate it by weighting each possible result by its chance and summing. It is most informative for repeated or portfolio decisions; with few tries, realized results can be far from the average.

In venture investing, backing many startups can make sense when a few large wins outweigh many smaller losses across the portfolio.

In healthcare, a screening program is attractive when the average benefit across patients exceeds expected costs and harms.

Insurers set premiums so expected payouts and expenses stay below expected revenue to support solvency, while managing tail risk and capital constraints.

For one‑off, high‑stakes choices, weigh downside, variability, and your own risk tolerance in addition to expected value.

Expected value anchors probability‑based decision models; behavioral and engineering work shows when averaging alone is not enough.

• Expected utility theory formalizes when maximizing a weighted average of utility is rational, building on early insights and axioms for choice under risk [2]. (1944)
• Prospect theory and subsequent reviews document loss aversion and probability weighting, explaining systematic departures from pure expected‑value choice [4]. (2013)
• Risk‑aware methods in machine learning, such as safe reinforcement learning, aim for high average returns while managing tail risk in practice [5]. (2015)

Expected value is informative but not sufficient on its own.

• It collapses outcomes into one number. For non‑monetary stakes (for example, safety or health), averaging can hide ethical trade‑offs.
• EV is a theoretical long‑run average; in one‑shot or few‑shot decisions—especially when ruin is possible—the distribution (worst case, tail risk) and constraints matter as much as the mean.
• People often misjudge probabilities and weigh losses more than comparable gains, so observed choices can diverge from EV benchmarks.
• Equal expected values can mask very different risk profiles. EV is the probability‑weighted average; risk refers to how spread out outcomes are around that average (for example, variance and tail outcomes).
• Dollars are not the same as utility or welfare. Decisions typically aim to maximize expected utility or overall value, not just expected accounting payoff.