Continuous Forecast Revision Loop
Update forecasts whenever new evidence strengthens or weakens the case
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 98%
The loop treats a forecast as a living probability assessment rather than a public commitment that must be defended. First, state the expected outcome, assumptions, and causal mechanism. As new information arrives, deliberately ask what supports the view and what contradicts it. Update the probability and implications rather than waiting until a distant deadline to declare the original call right or wrong. If the expected event repeatedly fails to occur or its mechanism no longer holds, revise the forecast and change the decision. The method reduces attachment to identity and reputation by making revision part of the process from the beginning. Its output is not perfect prediction but a better-calibrated current view that stays responsive to evidence and limits the cost of remaining confidently wrong.
Origin
Extracted from Masters in Business during Torsten Slok's discussion of Philip Tetlock's Superforecasting and the need for investors and sell-side analysts to revise views continually.
Core principles
- 01Treat forecasts as provisional beliefs
- 02Search for disconfirming evidence every day
- 03Separate reputation from the current view
- 04Change the forecast when the mechanism fails
- 05Judge reasoning continuously rather than only at the deadline
How to run it
- 1
Make the forecast explicit
Record the expected outcome, time horizon, probability, and assumptions supporting it. Define the mechanism that should produce the outcome.
Pro tip Use a probability rather than only a categorical prediction.
Watch out A vague forecast cannot be meaningfully updated or falsified.
- 2
Define revision signals
Specify which observations would strengthen, weaken, or invalidate the forecast before those observations arrive.
Pro tip Include at least one signal that would force a change of mind.
Watch out Do not choose only evidence that can confirm the existing view.
- 3
Review new information
At a regular cadence, ask what arrived that supports the forecast and what arrived that contradicts it.
Pro tip Review both sides in the same session to reduce confirmation bias.
Watch out Frequent review should not become overreaction to meaningless noise.
- 4
Update the probability
Adjust confidence in proportion to the diagnostic value of the new evidence. Preserve a record of what changed and why.
Pro tip Make small updates for weak evidence and larger updates for evidence tied directly to the causal mechanism.
Watch out Do not retain the same probability while merely rewriting the explanation.
- 5
Revise the decision
Change the forecast, position, or plan when the accumulated evidence no longer supports the original view.
Pro tip Treat a timely revision as successful forecasting behavior, not as an admission of personal failure.
Watch out Repeating that an outcome 'has gotta' happen is not evidence.
In the wild
An analyst forecasts rising equities based on strong earnings and fiscal support. Each day, the analyst reviews earnings revisions, credit conditions, policy changes, and evidence of trade-related investment delays. Several weak signals cause small probability reductions; a clear deterioration in the original earnings mechanism triggers a larger revision and a more defensive position.
→ The portfolio responds to changing evidence before the original annual forecast expires.
Common mistakes
Defending the public call
Forecasters may protect their reputation by interpreting every development as support for the original prediction.
Updating on noise
Changing the view after every weak data point creates instability rather than calibration.
Waiting for final failure
A forecast should be revised as its probability changes, not only after the predicted deadline has passed.
Is it for you?
Best for
Investors, analysts, and leaders making repeated decisions under changing conditions.
Not ideal for
Decisions governed by fixed rules where new information cannot alter the permitted action.
From the transcript
“you gotta update your prize and update your forecast constantly”
“I gotta have constant evaluation, constant thinking every day, what is the reasons why I'm right? What are the reasons that have come in today…”
“If it's not happening, you've gotta back down and say this is not happening.”
From the episode
Torsten Slok Discusses the Complexity of Global Economics
Torsten Slok