Predict the Mistake, Not the Outcome
Build an edge by forecasting systematic errors instead of unknowable fundamentals
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 95%
This framework shifts forecasting away from the difficult task of predicting fundamentals better than every large competitor. The practitioner starts with a well-documented bias, defines how that bias systematically distorts judgments, and searches for situations where the distortion is observable. A portfolio can then target overreaction or underreaction without claiming superior earnings forecasts. The mechanism resembles predicting that a sinker will produce ground balls: the forecaster need not hit the pitch or know the final score to anticipate a recurring response. Discipline is essential because every strategy must remain tied to a specific behavioral mechanism, tested against data, and protected from discretionary forecasts that lie outside the team's demonstrated advantage.
Origin
Richard Thaler described the behavioral strategy used at Fuller & Thaler Asset Management, where strategies target market mistakes rather than attempting to outforecast larger firms on earnings.
Core principles
- 01A deviation is useful only when its direction is predictable.
- 02Competing directly on crowded forecasts may offer no edge.
- 03Base each strategy on a specific, evidenced behavioral bias.
- 04Separate forecasting other people's errors from forecasting fundamentals.
How to run it
- 1
Select a documented bias
Choose a behavioral tendency with robust evidence, such as overreaction, underreaction, limited attention, or loss aversion.
Pro tip Prefer large effects that have survived adversarial replication and field testing.
Watch out Do not build a strategy around a clever but fragile laboratory result.
- 2
Define the predicted error
State exactly how the bias should push beliefs, prices, or decisions in the wrong direction.
Pro tip Write the prediction before examining the outcome data.
Watch out Saying that a price is simply wrong is not actionable.
- 3
Locate triggering conditions
Identify events or environments where the bias is likely to activate and where its effects can be measured.
Pro tip Use observable events such as surprising news, neglected information, or attention shocks.
Watch out Avoid retrofitting a behavioral explanation to every past winner.
- 4
Test persistence and economics
Evaluate the predicted error across time periods, populations, and realistic transaction costs.
Pro tip Include out-of-sample tests and comparisons with plausible alternative explanations.
Watch out Statistical significance does not guarantee an investable advantage.
- 5
Enforce the boundary
Restrict decisions to the validated behavioral edge and prohibit unsupported forecasting outside it.
Pro tip Create explicit process rules that separate permitted signals from discretionary stories.
Watch out Overconfidence can turn a narrow edge into an unjustified claim of general forecasting skill.
In the wild
An investment team identifies situations where investors systematically overreact to salient negative news. It defines the expected pricing error, tests whether prices subsequently mean-revert, and buys only when the documented trigger and valuation conditions are present. The team does not rely on its own point estimate of next quarter's earnings.
→ The strategy competes on predicting a recurring market mistake rather than on crowded earnings forecasts.
Common mistakes
Calling every disagreement a bias
A useful behavioral thesis must predict the direction and conditions of an error, not merely claim that the market is mistaken.
Forecasting outside the edge
Once a behavioral strategy is established, discretionary fundamental predictions can reintroduce the overconfidence it was designed to avoid.
Ignoring changing conditions
A historically persistent bias may become less profitable after costs, competition, or institutional changes.
Is it for you?
Best for
It is best for investors and strategists who can identify repeatable, measurable errors made by other participants.
Not ideal for
It is not ideal when the proposed bias has weak evidence, tiny effects, or no practical way to observe its triggers.
From the transcript
“we try to find stocks that we think the rest of the market is making a mistake about,”
“we forbid the portfolio managers from forecasting earnings”
“We're trying to predict the mistakes.”
From the episode
How Investors Fall Into Bias Traps with Economists Richard Thaler & Alex Imas
Alex Imas