How to use it 01 Start from a decision node and add one branch per option; each branch can carry a cash flow (negative for a cost) that is paid when that branch is taken. 02 Add chance nodes with branches for the possible outcomes and give every branch a probability. 03 End each path in a terminal node with its payoff. 04 Read the root EMV and the strategy table, which names the chosen branch at each decision node.
Worked readouts Two trees, both from the panel's own arithmetic. The demo drill decision spends 500 on the branch, then hits a 25% chance of a 2,400 payoff:
Drill branch: −500 + 0.25 × 2,400 + 0.75 × 0 = 100 ; staying out is 0, so the root EMV is 100 and the strategy table says "Drill".
A capacity choice: build big costs 400 with demand 30% high (1,500) and 70% low (200) → −400 + 0.3 × 1,500 + 0.7 × 200 = 190 ; build small costs 150 → −150 + 0.3 × 700 + 0.7 × 100 = 130 ; do nothing is 0. The rollback picks Build big with EMV 190 .
The drill's break-even success probability is 500/2,400 = 20.8% : below that, drilling's EMV turns negative and the optimal branch flips to staying out.
Backward induction Rollback starts at the leaves and moves left. A terminal contributes its payoff; a chance node contributes the probability-weighted sum of its branch totals, where each branch total is its own cash flow plus the child's rollup; a decision node contributes the largest branch total and records the winning branch. The root result is an expected value under risk neutrality — it treats a coin flip for 1,000 and a certain 500 as equally attractive when both average 500. The engine refuses to produce an EMV when any chance node's probabilities do not sum to 1: a node summing to 0.7 would make the weighted total meaningless, so the panel shows a labelled "draft total" (0.5 × 100 + 0.2 × (−50) = 40 in the example) with no optimality marking at all.
Limits
Probabilities are judgements. The engine checks that they sum to 1, not that they are calibrated; an overconfident 0.9 is still an overconfident 0.9.
Risk neutrality. EMV is the right yardstick for repeated, affordable bets. A single bet large enough to matter — a company-ending loss — deserves a utility or survival constraint that this panel does not model.
No discounting, no learning. All payoffs sit in the same period, and the tree cannot observe an outcome and choose again unless you build a second decision node. What it does not do. It does not estimate probabilities from data, sensitivity-test them for you, or produce a distribution of outcomes; for a range of results use the scenario Monte Carlo . This page provides general information and scenario estimates based only on the inputs you provide. It is not investment, tax, legal, or other professional advice and does not recommend any rate, product, or institution.
Sunk costs belong on the branch The most common modelling error is subtracting the cost twice or not at all. The panel expects the cost on the branch that incurs it — the drill branch carries −500 and the terminal carries the gross 2,400 — so the engine never asks you to convert gross to net by hand. The second trap is comparing top-line outcomes: 2,400 beats 500, but multiplied by a 25% chance it does not beat a certain 0 forever; at a 10% success probability the drill branch is worth −260 and the tree keeps you out. EMV is a long-run average, not a promise about the single decision in front of you.
Where it fits A go/no-go with real downside A product launch, an exploration well or a pilot line: list the up-front cost on the branch, the chance outcomes with their own probabilities, and read which option the rollback survives.
Comparing a certain and an uncertain option Staying out is a certain 0; the drill branch is 100 on average. The strategy table makes the comparison explicit, and the break-even probability tells you how confident the uncertain option has to be before it wins.
Privacy The tree, probabilities and rollback results are computed in your browser; nothing you type is uploaded.
References
Wikipedia, Decision tree , en.wikipedia.org (访问日期:2026-10-07)— decision and chance nodes, rollback.
Wikipedia, Decision analysis , en.wikipedia.org (访问日期:2026-10-07)— expected value and risk attitude in decision problems.
Wikipedia, Expected value , en.wikipedia.org (访问日期:2026-10-07)— the probability-weighted average.
OpenStax, Principles of Finance , openstax.org (访问日期:2026-10-07)— capital-allocation decisions under uncertainty.
Calculators in this hub
Hand-picked tools, one click away. The mini versions compute live and carry your values into the full calculator.
Sources & review
Reviewed by CalcX Editorial Team
Updated 2026-10-07