Forecasting Methods

Read these first

1. 01Calibrating a Forecast: Are Your Numbers Real?

   A model that says '+2% expected return' should, on average across every time it said that, be right around +2%. Checking whether it actually is, calibration, catches a whole class of quietly broken forecasts that a Sharpe ratio alone will not.

2. 02Shrinking a Forecast Toward Zero

   Your raw forecast is almost always too extreme. Pulling it partway toward zero, shrinkage, sounds like giving up on your own model, but it's usually the single biggest improvement you can make to a forecast's real-world accuracy.

3. 03The Grinold Alpha Formula

   Grinold's alpha formula is the bridge between a signal score and an expected return in real units: expected return equals the information coefficient, times volatility, times the standardised score. It is the single most-used formula in a research seat.

4. 04Turning a Score Into an Expected Return

   A model score tells you a stock ranks well. It does not, by itself, tell you how much return to expect, and the mapping from one to the other is a modelling choice with real consequences, not an afterthought.

Then the rest

1. 05Choosing the Shrinkage Intensity for a Forecast
2. 06Choosing What to Forecast
3. 07Classification vs Regression Targets for Trading
4. 08Combining Forecasts Across Horizons
5. 09Conditional Forecasts on State Variables
6. 10Forecasting Correlation, Not Just Volatility
7. 11Forecasting Direction vs Magnitude
8. 12Forecasting Returns vs Forecasting Risk
9. 13Forecasting Through a Regime Change
10. 14Forecasting Your Own Trading Costs
11. 15From Expected Return to Position Size
12. 16Measuring the Economic Value of a Forecast
13. 17Picking a Forecast Horizon
14. 18Quantile Forecasts and Asymmetric Loss
15. 19Raw, Excess or Residual Return as the Target
16. 20Regressing Realised on Predicted
17. 21Reliability Diagrams for Trading Forecasts
18. 22Sizing Down When the Forecast Is Uncertain
19. 23Volatility Forecasts as a Sizing Input
20. 24Why a 1% R-Squared Can Be a Great Forecast