To be clear I'm using the code from Richard Santos for the functional polynomial regression. I really loved his script idea (given that I'm into linear regression myself). I took his code and made a strategy which applies to bitcoin on the 5 minute chart (but you could adjust this for any asset), and you could make this work on anytime frame by adjusting the 'length' property of the regression until you get good results.
This strategy is very simple.
I drew lines to represent the bottom most part of the regression (green line), and top most part (red line). If the close crosses under the red line then you short. The reverse if it crosses above the green line then you go long.
Very simple but effective. To understand this script be sure to add Richard santos function polynomial regression so you can see what is going on and what I'm trying to do.
For example, on BTC with a length of 61 if you bought and held you would have only had 5% gain, instead with this strategy you are looking at very close to 60% gains! Now this could be a weird fluke of over-optimizing but I'm not sure this is the case because you can adjust the length from there +/- about 10 and still have good results. It's just certain lengths are the correct trading rhythms for different assets.
I'm excited to see how you guys use this and if you have any success. Be sure to thank Richard Santos for his great work!