# Many Worlds

Source: https://dinand.com/experiments/many-worlds/
Author: Dinand Tinholt

> Scenario planning that plays a business decision out across hundreds of thousands of simulated futures and tests every response against the same draws.

Recommended responseRefining

## Running the first simulation.

Reasoning

## Why this is the call

Robustness

## How sturdy the call is

Plan

## The plan, quarter by quarter

Committed moves start this quarter, while an option gets prepared now and is exercised only if its trigger fires.

Moves

## Value of each move

Each bar shows how far the risk-adjusted score moves when you add a move to the recommended set or take it out. Every comparison runs on the same futures.

Outcomes

## Where the outcomes land

Failure modes

## When the call is wrong

The engine searches the futures for the conditions that best explain where the recommendation loses.

Stress tests

## How the call fares when assumptions are wrong

Exposure

## Exposure that remains

The average outcome under the recommended response when each uncertainty sits at either end of its range.

Budget

## Return on more budget

The best score the search found at each level of upfront spend.

Search

## Every policy tested

Each dot is one combination of moves, committed or held as options. A dot sits higher when its expected value is larger and further right when its worst tenth of futures hurts less.

Finalists

## The finalists, side by side

Analyst

## Ask Claude

Claude reads the scenario and the results, and reruns the simulation itself to test a what-if. It uses your own Claude account.

Inputs

## Assumptions

The engine uses every input listed here, and editing any value reruns the simulation. Each range runs from a plausible low to a plausible high, with its peak at the most likely value.

### Fit ranges and correlations from data

Paste a table with a header row and one column per driver, in the driver's own unit such as yearly % change. Each fitted range spans the 5th to the 95th percentile and peaks at the median. Correlations come from the matched columns.

---
Site guide for AI agents: https://dinand.com/llms.txt