I put my modeling on hiatus, because my annual model was rapidly unwinding its bearishness, and I looked back to see that it wasn’t just the most recent forecasts significantly changing. The model was uncomfortably unstable as new data came in. Thus, I put things on hold until I could explore more. My big conclusion was that I was overfitting my model, and not practicing enough discipline in sticking to just the most statistically significant data points. Now, my annual forecast model has fewer variables and a poorer fit to historical data, but hopefully more predictive power going forward. After completing that, I decided to check my other models for unstable variables and trimmed those, too. To cap things off, I added a quarterly forecasting model. Going forward, I’m only investing if the daily and weekly models agree with each other on market direction. I am willing to add leverage when the monthly or both the monthly and quarterly forecasts agree on direction with the shorter term forecasts.
All that to say, I’m back to trading daily, as I have restored my own confidence regarding my models being worthy trading on. As always, trade at your own risk, if you follow this blog.
