skills.rest
radarist
radaristOFFICIAL
Explore all skills by this author →
WebsiteGitHub

expected-value-decision-tree

Rolls back decision trees to expected values and prices information with EVPI, EVSI, and tornado sensitivity.

4|1|Updated Jul 30, 2026
Search Tags:#decision tree#decision analysis#sensitivity analysis#monte carlo#expected value#evpi#evsi
One-click install
npx skills add https://github.com/radarist/structured-analytic-skills --skill expected-value-decision-tree-radarist
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: expected-value-decision-tree
Source: https://github.com/radarist/structured-analytic-skills/tree/main/skills/expected-value-decision-tree
Command: npx skills add https://github.com/radarist/structured-analytic-skills --skill expected-value-decision-tree-radarist

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Staged bets like go/no-go investments or pilot-then-commit choices hinge on uncertain outcomes, and it is hard to know whether a pilot or study is worth its cost before committing. This Skill lays out the decision as a tree, rolls it back to expected values, and computes exactly what perfect or sample information is worth. ## Core Features & Use Cases - Decision tree roll-back: Computes expected values per option, the optimal policy including contingent choices, and the risk profile (probability of loss, best/worst outcomes). - Value of information: Calculates EVPI and EVSI against a test's cost with an explicit run/skip verdict, so you only pay for pilots that can change the decision. - Sensitivity analysis: Produces a one-way tornado ranking with switching points, plus seeded Monte Carlo over input ranges for P10/P50/P90 distributions. - Use Case: Deciding whether to run a $2M pilot before a $20M development commitment — the Skill computes each option's EV, prices the pilot's information at EVSI 3.8 vs cost 2.0 (run it), and identifies that the recommendation switches if P(High demand) exceeds 0.43. ## Quick Start Ask the AI to build a decision tree for your go/no-go choice with probabilities and payoffs, and to report the expected values, EVPI, EVSI of any proposed test, and the sensitivity switching points.

Frequently Asked Questions about expected-value-decision-tree

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I calculate expected value of perfect information (EVPI) for a decision?▼

EVPI equals the expected value with perfect information minus the best expected value without it. Fix each outcome of the uncertainty, re-optimize the decision, weight by the prior probabilities, and subtract the base EV. The dtree.py script automates this with the evpi command.

How do I decide if a pilot or market study is worth its cost?▼

Compute the test's EVSI (expected value of sample information) from its likelihood matrix and compare it to the test's cost. Run the test only when EVSI exceeds cost; EVSI near zero means the result cannot change your decision. EVSI is always bounded above by EVPI.

What is a tornado diagram in decision analysis sensitivity?▼

A tornado diagram ranks every probability, payoff, and cost by how much a ±20% change swings the decision's expected value. It also identifies switching points — the input values where the recommended option flips — showing where estimation effort should go.

When should I not use a decision tree for a business choice?▼

Skip the tree when there is only one option, when no outcome would change the choice, or when options are ranked on multiple criteria with no material chance node (use multi-criteria decision analysis instead). If probabilities are unknown, elicit them first via Delphi or forecasting methods.

Does the dtree.py script require external Python packages?▼

No, dtree.py uses only the Python standard library and runs on Python 3.9 or later. It is deterministic with seeded randomness for Monte Carlo, and includes a --selftest mode that verifies results against the classic oil-wildcatter textbook tree.

Related Skills

Explore More Product & Management Skills→
AndurilCodeAndurilCode
7

probabilistic-thinking

Convert vague claims into Bayesian probability estimates with base rates and expected values.

Official
Intermediate
duany049duany049

exp-pilot-run

Execute pilot experiments from YAML specs and collect raw JSON results.

Community
Intermediate
jmdueajmduea

subagent-driven-development

Dispatch fresh implementer subagents to execute development plan tasks with sequential spec and code quality reviews.

Community
Advanced
calvyntwhcalvyntwh

decision-matrix

Score options against weighted user-defined criteria to quantify subjective decisions.

Community
Advanced