A very interesting topic, I would love to have arrays, multitimeframe, bigger loop capacity (the “infinite loop” error message arrives way too soon even when the loop is limited), ability to store patterns with a save function not to recalculate them at each session, and ability to draw in the future in order to project predictions from machine learning on pattern recognition and monitor reality against the baseline of the prediction, all within PRT… Prt dev team, if you read us, yes we’ve been told multitimeframe arrives soon, and thanks for that, but we’d love to have all the rest too in future versions to do this, thanks 🙂

Nice Topic! I am not use to the idea of machine learning but I will think a lot about it

About the first part of you code, can you please explain me how or why you define the variable “best” in this way and not another?

When I was reading the first post of this Forum “Build a smarter pattern matching algorithm that relies on high/low points instead of brute-forcing and checking every bar in every sequence. Does anyone have any ideas on this?”, something come to my mind:

What if instead of bars we study the movement of a Hull Moving Average everytime it changes the slope direction and we store the barindex and the level of the curve, so when we study the past we try to predict how will be the next direction change…