What is probabilistic forecasting in investing?
Reviewed against the platform's code on Sep 23, 2026
Probabilistic forecasting predicts a distribution of possible outcomes — for example a median return with a calibrated interval around it — rather than a single number, so the forecast states how uncertain it is and that uncertainty can be measured and used to size a position.
Why it matters
A forecast without a stated uncertainty cannot be sized: a +2% view held with high confidence and one that is barely distinguishable from noise call for very different positions.
How it works
Common methods include quantile regression, ensembles, and conformal prediction, which calibrates intervals on held-out data so that, for example, a 90% interval contains the outcome about 90% of the time. Calibration is then checked continuously.
How Opulence Alpha applies it
Opulence Alpha publishes forecasts as distributions over several horizons, with intervals built by conformalised quantile regression, and combines its model voices with online aggregation re-weighted on realised skill.
Today's fair-value board →Related concepts
Questions
- What is forecast calibration?
- Whether predicted probabilities match outcomes: events forecast at 70% should happen about 70% of the time.
- What is conformalised quantile regression?
- A method that fits quantile models and then adjusts their intervals on held-out data so they reach their stated coverage (Romano, Patterson & Candès, 2019).
- Why isn't prediction accuracy enough?
- Because accuracy says nothing about uncertainty or about how a forecast behaves in different market regimes, which is what position sizing needs.
References
- Romano, Y., Patterson, E. & Candès, E. (2019). Conformalized Quantile Regression. NeurIPS.
- Wintenberger, O. (2017). Optimal learning with Bernstein Online Aggregation. Machine Learning.
Educational content about research methods. Not investment advice.