Probability-Weighted Scenario Overlays
Layer correlated scenarios over a base model to price risk explicitly.
A bottom-up model produces one number; the world produces many possible futures. Probability-weighted scenario overlays sit on top of the base model and run multiple simulations across plausible futures — robotic production, regulatory shifts, macro shocks, technological disruption. Each scenario carries a probability and adjusts specific drivers in the base model.
Crucially, scenarios are correlated. If robotic factories arrive, labor cost assumptions collapse and capacity assumptions rise — the model has to encode those joint moves. Other scenarios are independent (a tech jump and a macro recession can co-occur). Running thousands of simulations across the joint distribution outputs a fair-value field, not a fair-value point.
The output looks like ARK's Monte Carlo distributions: a probability mass spike around the central estimate, with tails. If today's price sits well below the spike, that's a buy with a quantified margin of safety. If the price is at the spike but the left tail is fat, the risk-adjusted upside may not justify holding.
- A point estimate hides uncertainty; a distribution exposes it.
- Scenarios must encode correlations, not just isolated probabilities.
- Low-probability high-impact scenarios still belong in the model — they shape the tails.
- The fair-value output is a field, with a central spike and explicit tails.
- Risk is not the standard deviation of returns; it is the probability mass of bad outcomes.
- Start from a finished base modelScenarios overlay a bottom-up valuation. Without a working base model the scenarios have nothing to perturb. Lock the base before adding scenarios.
- Enumerate plausible scenariosList technological, regulatory, competitive, and macro scenarios specific to the business — for Tesla, robotic production lines, EV regulatory backlash, unionization, new entrants. Don't restrict to good or bad scenarios.
- Assign probabilities and impactsFor each scenario, assign a probability and identify which model drivers it moves. Robotic factories cut labor cost and raise capacity by 3-4x; regulatory backlash cuts addressable market.
- Encode correlations between scenariosSome scenarios force or preclude others. If robots arrive, labor inflation matters less. If a recession hits, capex plans get cut. Build a correlation matrix or conditional rules.Pro tipGet the obvious correlations right — perfect calibration isn't the point, avoiding double-counting is.
- Run a Monte Carlo simulationSample thousands of paths across the joint scenario distribution and compute fair value for each. Output a histogram of fair values.
- Compare today's price to the distributionIf price sits in the lower tail, you have margin of safety. If it sits at the spike with a heavy left tail, the risk-adjusted upside is poor. Decide accordingly.
Sasha includes a low-probability scenario where robotic production lines come online by 2030, slashing labor costs and raising capacity to 3-4x. It carries low probability but high impact — perturbing the right tail of the distribution.
ARK Invest publishes Monte Carlo simulations on names like Tesla, showing the spread of outcomes rather than a single price target. Sasha uses this as a public reference for the technique.
Sasha imported scenario modeling from his M&A consulting work, where pricing an acquisition required quantifying regulatory, integration, and synergy risks rather than picking one growth rate. He references ARK Invest's published Monte Carlo simulations as a familiar public example of the same technique.