Expected Value
Expected Value: Multiply each possible outcome by its probability, then sum across all outcomes. The result is the average return you'd receive if you made this decision many times — your rational basis for comparing any uncertain alternatives.
What Is Expected Value?
Expected value (EV) is a concept from probability theory that answers a fundamental question: if a situation with uncertain outcomes were repeated many times, what would the average outcome be?
The formula is straightforward: EV = Σ(probability of outcome × value of outcome) — summed across all possible outcomes.
For a coin flip where heads wins $100 and tails loses $50: EV = (0.5 × $100) + (0.5 × -$50) = $50 - $25 = $25 positive EV. If you played this game repeatedly, you would on average win $25 per flip. Any rational decision-maker should play this game.
The power of expected value is not that it predicts any individual outcome — it can't. It predicts the long-run average, which is what matters for decisions that repeat or for a portfolio of similar decisions. A single coin flip might lose $50. A thousand coin flips at this EV would produce approximately $25,000 in profit.
EV thinking is what separates professional gamblers from recreational ones, skilled investors from lucky ones, and systematic decision-makers from intuitive ones. It converts the question "will this work?" — unanswerable under uncertainty — into the question "does this decision have positive expected value?" — answerable with honest probability assessment.
The practical challenge is not the formula but the inputs: most probabilities in real decisions are not precisely known. This requires calibrated estimation under uncertainty — a skill developed through deliberate practice, reference class data, and honest accounting of what you don't know.
How It Works
Step 1: Enumerate possible outcomes
Identify the meaningful scenarios, not every possible state.
For most decisions, 3–5 scenarios capture the distribution.
Step 2: Estimate probabilities for each outcome
— Use historical base rates when available.
— Consult reference classes (what happened in comparable
situations?) rather than relying on the inside view.
— Probabilities must sum to 100%.
Step 3: Estimate the value of each outcome
— Express in comparable units (money, time, strategic value).
— Include downside outcomes with negative values.
— Be honest about the realistic worst case.
Step 4: Calculate EV
EV = (p₁ × v₁) + (p₂ × v₂) + ... + (pₙ × vₙ)
Step 5: Compare against alternatives
— The decision with the highest EV is rational to prefer.
— Adjust for risk tolerance (variance matters, not just EV)
for decisions involving your survival or ruin.
Example:
Startup investment: $100K check
Scenarios:
- 60% chance: lose the full $100K = -$60K weighted
- 25% chance: return 2x ($200K) = +$50K weighted
- 10% chance: return 10x ($1M) = +$100K weighted
- 5% chance: return 50x ($5M) = +$250K weighted
EV = -$60K + $50K + $100K + $250K = +$340K
This is a positive EV investment if your probability
estimates are correct.
Real-World Examples
Example 1: Poker — The Foundation of Professional Play
Professional poker players make every significant decision using expected value calculation. A player faces a pot of $1,000 with $500 remaining to call an opponent's bet. She estimates she has a 40% probability of winning the pot if she calls.
EV of calling: (0.4 × $1,500 to win) - (0.6 × $500 to call) = $600 - $300 = +$300.
This is a positive EV call — regardless of whether she wins or loses this particular hand, making this call repeatedly will produce positive results over time. This is the calculation professional players make for every significant decision, every hand. The discipline of acting on EV rather than emotion or fear is the primary differentiator between professional and recreational players.
The same logic applies to bluffing: a bluff is correct when the expected value of the bluff (probability of success × pot won) exceeds the cost (probability of failure × bet lost). EV calculation makes the decision mechanical rather than intuitive.
Example 2: Pharmaceutical R&D Investment
A pharmaceutical company is deciding whether to advance a drug candidate into Phase 2 clinical trials, which will cost $50 million. Historical data on drugs at this stage of development:
- 40% probability: Phase 2 fails (loss of $50M)
- 35% probability: Phase 2 succeeds, Phase 3 fails (additional $150M invested, total loss $200M)
- 20% probability: Phase 3 succeeds but commercial launch underperforms (NPV of $300M, less $200M total investment = $100M net)
- 5% probability: Phase 3 succeeds with strong commercial performance (NPV of $2B, less $200M = $1.8B net)
EV = (0.40 × -$50M) + (0.35 × -$200M) + (0.20 × $100M) + (0.05 × $1.8B) = -$20M - $70M + $20M + $90M = +$20M positive EV
The investment has positive expected value. The company should advance — but knowing the distribution also tells them that in 75% of cases, they will lose money. The positive EV is driven by the 5% extreme upside. This is why pharmaceutical R&D requires portfolio thinking: any single drug candidate is a loser in most worlds; a diversified portfolio of positive-EV bets produces reliable returns.
Example 3: Negotiation Decision
A freelance consultant is in a contract negotiation. She can either accept the current offer of $15,000, or counter with $20,000. If she counters:
- 50% probability: client accepts $20,000 (gain $5,000 vs. accepting now)
- 30% probability: client negotiates to $17,500 (gain $2,500 vs. accepting now)
- 20% probability: client walks away (lose the entire $15,000 contract)
EV of countering vs. accepting = (0.5 × $5,000) + (0.3 × $2,500) + (0.2 × -$15,000) = $2,500 + $750 - $3,000 = $250 positive EV
The expected value of countering is marginally positive. But the distribution matters here: there's a 20% chance of losing $15,000 — which may not be acceptable given her current financial situation. EV calculation says counter; risk tolerance and variance may say accept. This is where the Kelly Criterion and risk of ruin considerations modify pure EV thinking.
When to Use It
✅ For any repeating decision or portfolio of decisions. EV is most useful when decisions recur or when you're making many similar bets — the long-run average is what matters.
✅ When comparing options with different probability/payoff profiles. A 90% chance at $10 and a 10% chance at $80 are not self-evidently comparable — EV resolves it ($9 vs. $8; take the 90% option).
✅ When assessing whether a course of action is worth pursuing at all. Negative EV decisions should generally be declined regardless of how appealing the upside looks.
✅ When building investment or resource allocation frameworks that must perform consistently over time.
❌ For true one-time decisions where variance dominates EV. If a decision could ruin you — take your life savings and bet them on a positive-EV outcome — the EV calculation is insufficient without also accounting for the survival constraint.
❌ When probabilities are completely unknowable. EV requires probability estimates. In genuinely novel situations with no reference class, any probability estimate is speculation, and the EV calculation inherits that uncertainty.
Model Combinations:
| Combine with | Effect |
|---|---|
| Bayesian Thinking | Bayesian thinking calibrates the probability estimates that feed into EV |
| Kelly Criterion | Kelly extends EV to determine the optimal bet size given your bankroll |
| Margin of Safety | Apply margin of safety to your probability estimates before calculating EV |
Common Misuses and Limitations
Misuse 1: Overconfident probability estimation. The EV calculation is only as reliable as its probability inputs. Most people systematically overestimate success probabilities and underestimate failure probabilities. Using base rates and reference class forecasting is essential.
Misuse 2: Ignoring variance in single-occurrence decisions. For decisions where the outcome is existential — bankruptcy, ruin, irreversible harm — pure EV maximization is insufficient. You must also consider the probability of catastrophic outcomes independently of EV.
Misuse 3: Garbage-in, garbage-out precision. A neatly formatted EV table with highly precise-looking numbers creates false confidence if the underlying probability estimates are guesses. Acknowledge the uncertainty range in your inputs and express EV as a range, not a point estimate.
Related Models
Kelly Criterion: The mathematical extension of EV that determines the optimal fraction of capital to allocate to a positive-EV bet.
Bayesian Thinking: Provides the framework for updating probability estimates — the inputs to any EV calculation — as new evidence arrives.
Asymmetric Risk: Applies EV thinking specifically to situations where the upside is disproportionately large relative to the downside.
FAQ
What if I can't estimate the probabilities precisely?
Imprecise probability estimates are still useful. Express your uncertainty as ranges: 'I think the probability of success is between 20% and 40%' produces an EV range rather than a point estimate. This is more honest and often more useful than false precision. For most real decisions, the sign of the EV (positive or negative) is more important than its exact magnitude, and rough probability estimates are typically sufficient to determine the sign.
When does EV break down as a decision framework?
EV breaks down when (1) the decision is a true once-in-a-lifetime event where the long-run average is irrelevant, (2) the variance is so large that it matters as much as the mean (a 1% chance of ruin alongside 99% small gains may be negative EV in a meaningful sense even if the formula says positive), and (3) probabilities are genuinely unknowable — when there is no meaningful reference class at all.
What's the best resource for learning Expected Value?
For the formal foundations, any introductory probability textbook covers EV rigorously — Sheldon Ross's A First Course in Probability is standard. For applied decision-making, Richard Zeckhauser's papers on 'knowns and unknowns' (search on SSRN) are excellent. For investing applications, Howard Marks's The Most Important Thing (2011) and Michael Mauboussin's More Than You Know (2006) cover EV thinking in detail.
Apply This Model with AI
Describe your decision and the possible outcomes you see in MindMax. The AI will help you structure the probability estimates, calculate the expected value, and compare it against alternatives — including flagging where your probability assumptions may need calibration.
🚀 Apply Expected Value in MindMax →
Further Reading
- Michael Mauboussin, More Than You Know (2006) — The clearest treatment of expected value in the investing context.
- Annie Duke, Thinking in Bets (2018) — Practical application of EV thinking to decision-making, drawing on her experience as a professional poker player.
- Howard Marks, The Most Important Thing (2011) — Chapter 7 covers EV thinking in the context of asymmetric risk and second-level thinking.
This page is part of the MindMax Mental Models Knowledge Base.