Hi

I surprised its not in the pro screener as it would add value. I look for them manually and sometimes trade them, so be interesting to see the back test results….

Jamie

Hi Grahal.

the issue is the double tops don’t look like a classic double top pattern as spotted by eye on a chart. My code is defining a fixed number of bars apart, but classic double tops are not a regular fixed number of bars apart.

Good 🙂 This is the point I was waiting to see in this post and why I commented above on writing it in plain English first:-)  Once you get coding you soon come across this hurdle. That’s then followed by some other things to solve for.

There are several ways to approach this. Each way has pros and cons. As neural networks are not part of PRT you will need to start with some rigid hard coded constants on with, depth, noise factors etc. This way is sadly much less effective.

The thing to do is decide upon a sensible range of with and height that your pattern could be.  On the surface this looks simple  (as do oblique trendlines) but programmatically it’s a challenge to get right. Your validation constraints should auto adjust depending on price action dynamics.

A double top is an M pattern. You  need to ask the question of how do you validate it, given that it could be any size. Rigid constants are not really the answer but help you get started. How does your brain do it? Can you put it into English let alone code?

Write down some rules in English and follow them and then see if it works on out if sample data.

This is actually a bigger topic than you might think.

I  worked on a head and shoulders based stop loss for PRT. Challenging to do as elegantly without learning systems