Skip to content
StockMarketAgent
Methodology · Scenarios

Scenario probability tree

A probability-weighted tree of mutually exclusive macro scenarios used to bound the index implied fair-value range with explicit conditional assumptions on growth, inflation, policy rate, and earnings.

Specification

Scenario probability tree — operational spec

Imposes the discipline that probabilities sum to 100% and each branch carries an explicit assumption set, so a wide spread between bull and bear cannot be used to hedge a directional view.

Inputs
  • Macro branch definitions (typically soft landing, hard landing, no landing, reaccelerating inflation)
  • Conditional growth, inflation, policy-rate, and earnings assumptions per branch
  • Branch-level fair-value model output (bottom-up index FV under each assumption set)
  • Branch probabilities — explicitly chosen, not implied from a model
Computation
  • Each branch is required to be mutually exclusive and the probabilities must sum to exactly 100%.
  • Branch fair value is computed using the same valuation model used for individual companies (discounted earnings at index Ke, normalized terminal P/E, with the macro assumption set fixed).
  • Implied fair value = sum over branches of probability times branch fair value.
  • Implied fair-value band = (worst branch FV, best branch FV).
  • Branches are recomputed every month as conditional probabilities are repriced.
Outputs
  • Probability-weighted implied fair value (single number).
  • Implied fair-value band (low, mid, high).
  • Branch probabilities, each with its conditional assumption set surfaced.
  • Sensitivity table showing how the headline FV moves under +/- 10pp probability shifts on each branch.
Limitations
  • Branch probabilities are subjective. The tree is auditable but not objective.
  • Mutually exclusive branches force a discrete view of a continuous macro state space; edge cases (stagflation that is half hard-landing, half reacceleration) are awkward.
  • Fair-value computation under each branch inherits the assumptions of the underlying valuation model, including terminal P/E and discount rate.

Frequently asked

Are the branch probabilities backed by a model?
No. The tree intentionally surfaces probabilities as explicit, challengeable inputs rather than model-derived numbers. Probability models trained on macro data are unstable across regimes, and laundering subjective probabilities through a model produces false objectivity.
Why mutually exclusive branches?
Mutually exclusive branches force the analyst to take a position. Overlapping branches let the writer hedge — assigning 30% to soft landing and 30% to no landing means 60% of the probability is on outcomes that share most of their assumptions, which is not really a view. Mutual exclusivity makes the spread between branches do real work.
How often is the tree updated?
Monthly, alongside the regime composite. Probabilities are repriced based on the prior month's macro data and policy signals. Branch definitions are kept stable across months unless the macro state space materially changes; rolling-the-tree should be a deliberate decision, not silent.
See the full methodology hub for the rest of the model registry, or open the glossary entry for the headline definition.