Say I have some features that I think have predictive power. I sample the data and visually it appears when the indicator reaches a certain threshold the stock price will usually move higher. What are the best ways to test if this relationship is true across all of the data I have? Usually this indicator occurs anywhere from 3-14 days before the price moves higher. I could write a test to record performance every time the indicator reaches a certain threshold but I'm wondering if there are any techniques I should look into to evaluate indicators, without iterating through conditions set by trial and error. The questions I have about the indicator is, what is the optimal threshold to trigger a buy, what is the average holding time in days that produces the best return, and does the indicator perform better or worse with another feature introduced.
Is this just something that is trial and error or are there techniques out there that can help determine the value of an indicator?
Submitted October 11, 2020 at 06:06PM by Eccs15
via https://ift.tt/3jR5jzD