Rules-Priors-Evidence Probability Method
Build coherent forecasts by updating historical baselines with new evidence
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 98%
The method begins by defining the event, its deadline, and the institutional rules that constrain possible outcomes. The analyst then reconstructs a historical base rate, such as how often governments have completed comparable policy priorities. New evidence is layered onto that prior probability, including party control, procedural options, internal agreement, delays, or failed precursor legislation. Bayes' theorem provides the conceptual mechanism: start with what normally happens and update the estimate as relevant facts arrive. The final percentage is not presented as certainty or false precision; it is a disciplined input for constructing a coherent portfolio story, comparing scenarios, and managing event risk. Independent colleagues can provide a final sanity check against overlooked assumptions.
Origin
Extracted from Masters in Business, where Michael Zezas explains how Morgan Stanley estimates the probability of policy events such as tax reform.
Core principles
- 01Treat probabilities as decision tools, not precise predictions
- 02Define the event and its constraints before estimating odds
- 03Anchor forecasts in historical base rates
- 04Update the baseline when relevant evidence changes
- 05Sanity-check estimates with independent thinkers
How to run it
- 1
Define the Event
State exactly what outcome is being forecast and establish the relevant deadline. Distinguish timely success from an outcome that arrives too late to satisfy the investment thesis.
Pro tip Write down what would and would not count as success before reviewing evidence.
Watch out A vague event definition allows the forecast criteria to change after the fact.
- 2
Map the Rules
Identify the legal, procedural, political, or market constraints that determine how far the outcome can move in either direction.
Pro tip Separate formal constraints from assumptions about participants' intentions.
Watch out Motivation alone does not overcome incompatible rules or coalitions.
- 3
Establish the Prior
Use historical data to estimate how frequently comparable events have succeeded. Select the closest defensible reference class rather than relying on intuition.
Pro tip Reconstruct the historical record when a usable dataset does not already exist.
Watch out An irrelevant reference class produces a misleading baseline.
- 4
Layer in New Evidence
Identify current facts that make success more or less likely than the historical baseline. Update the estimate in a consistent direction and magnitude.
Pro tip Record why each fact changes the probability so later revisions remain auditable.
Watch out Do not count the same evidence multiple times under different labels.
- 5
Sanity-Check the Estimate
Compare the result with Bayes' theorem, historical analogues, and the views of other informed analysts. Investigate meaningful disagreements rather than averaging them away.
Pro tip Ask colleagues to challenge the reference class and missing evidence.
Watch out Consensus can reinforce a shared blind spot.
- 6
Connect Odds to Decisions
Translate the probability and its alternative scenarios into portfolio exposures, timing risks, and potential losses. Revise the estimate when material evidence changes.
Pro tip Track both the probability of success and the probability of success within the required timeframe.
Watch out A correct long-run forecast can still produce a losing time-sensitive position.
In the wild
An analyst defines success as comprehensive tax reform within twelve months, maps the budget-reconciliation rules, and reviews postwar administrations to estimate how often presidents with unified party control execute top priorities. The analyst then adjusts for congressional control, reconciliation eligibility, party disagreement, and delays caused by failed healthcare legislation.
→ The portfolio receives a defensible probability estimate and can distinguish timely reform from delayed or disruptive variants.
A bank strategist starts with the historical passage rate of comparable regulations, identifies the agencies and legislative votes required, and updates the baseline for current political control, court challenges, and implementation delays. The resulting scenarios are mapped to affected securities rather than treated as a binary prediction.
→ Exposure is sized according to evidence-weighted scenarios instead of political conviction.
Common mistakes
Presenting False Precision
A detailed percentage can imply more certainty than the evidence supports. The estimate should communicate disciplined reasoning, not clairvoyance.
Ignoring the Timeline
An event that eventually occurs may still disappoint investors if it misses the market's expected window. Forecast timing as explicitly as the final outcome.
Confusing Intent With Ability
Strong political motivation does not ensure that procedural rules or internal factions can be reconciled.
Is it for you?
Best for
It is best for investors and strategists assessing event risks with historical precedents and changing evidence.
Not ideal for
It is not ideal when the event is undefined, unprecedented, or unsupported by relevant evidence.
From the transcript
“what probabilities, as long as they're well defensed and logical, will help you do as they help you build some logic into your portfolio, into…”
“we try and we whatever problem, whatever event we're trying to to to forecast here, we first try and establish what are the rules of…”
“Bayes' theorem is just a way of calculating the probability, or at least the way we use it, it's the way of calculating the probability…”
From the episode
Interview With Michael Zezas: Masters in Business (Audio)
Michael Zezas