Other important notes
In tools:
Correlation analysis threshold 0.98
Detect balance lines correlation: on
Detect strategies with similar trading rules: on
Trading session: open to all hours of every day
Data Horizon:
Maximum Data Bars: 200,000
Minimum Data Bars: 300
Use start date limit: 32 months ago from the day I generate
Use end date limit: 2 months ago from the day I generate
Validator 1:
I take all of the collections I just made and run them through these settings.
Acceptance criteria: Profit Factor 1.2
Out of Sample: In sample
No Normalization or Monte Carlo this time.
Use start date limit: 2 months ago from the day I generate
Use end date limit: 1 month ago from the day I generate
Validator 2:
I take all of the collections that passed Validator 1 and run them through these settings
Acceptance criteria: Profit Factor 1.2
Out of Sample: 50% OOS
No Normalization or Monte Carlo this time.
Use start date limit: 1 month ago from the day I generate
Use end date limit: the day I generate
Then I pick the top 10 strategies for each pair and export a portfolio.
I immediately place them onto the challenge account.
The idea here is to create strategies that profited for long periods of time and are still profitable for the last 2 months and profitable bi-weekly last month.
The performance is never as good as the backtest but that’s the whole idea behind forex I think. As I’ve heard from a user in Popov’s forum “You can’t train algebra.”
Anyways I run these strategies for exactly 2 weeks. I delete them, then create new strategies using the same generation settings.
By the way, I’m now doing this through Express generator cuz I can do almost the entire process with a single click of a button.
It was very tedious to do this before hand.