Did you know that any Random Forest program, and other modern programs, output “Feature Importance.” This is just the importance of factors and functions we use in our ranks.
De Prado considers this THE ONLY WAY TO DETERMINE WHICH FEATURES TO USE and says using backtest to determine which features are important is about the biggest mistake one could make. Is he right? I will not argue the theory here because you can get it directly from him in his book and decide for yourself. But there are some clear advantages to his way.
You can run a Random Forest on a boatload of features, get a cup of coffee, look at the output and know what works. Keep in mind that in addition to being easier de Prado says it is better.
Or not. Optimize for the rest of you life in you prefer.
Me, I don’t get out enough to spend my money the way it is. Plus, de Prado says I will have a little more money to spend when I do get out. Are we sure he is wrong?
-Jim