Question and little problem with the Data Parts

There’s something I don¡t understand about the Data Parts and their role in the creation strategy part and I’d love some light on this problem:

For what I’ve read, the strategies are created in the white part, where there is no ISV or OOS.

Then, the ISV (grey part) is the place where the previously created strategies are developed and refined.

And finally, the OOS (green part) is where strategies are tested, and this Data Time Range is no used for the strategy evolution.

Now my question is:

If I change those Data Ranges, will I get different strategies?

I’m having a problem where the Equity Chart of the strategies stagnates and it goes lineal, the profits and the losses are cancelled and I win no real money.

There’s a way to solve this using Data Regions?

I’ve been playing around and I’m getting the same results over and over.

Any help would be much apreciated!

And another fast question non-related wirth the other one: What do you think about using already generated strategies as Initial Generation? Does it work for you? I try to get some random strats, then use them as a fine tune for new strategies using the Initial Pop.

Strategies are basically created using in-sample (white) data. If you change the white data range, you will get different strategies, because this is where the logic and edge are discovered. The grey ISV and green OOS parts do not create or improve strategies — they only test and filter what was already built.

If your equity curve is flat, it usually means the strategy has a weak or diluted edge. This often happens when the white data range is too long or contains many different market regimes. StrategyQuant then tries to build ‘safe’ strategies that work a little everywhere but don’t make real money. A better approach is to use a shorter, more focused white range (for example 3–5 years or one volatility regime).

Data Regions are not meant to ‘fix’ a strategy. Their real value is to show when the strategy works and when it doesn’t. You can then trade the strategy only in the conditions where it has an edge (for example high volatility or specific session times), instead of forcing it to work all the time.

As for the other question: Yes, using existing strategies as the initial population is a perfectly valid approach and gives you better control over the development process.

I see that, thanks Tomas!

As I cannot change the white part, just add the other ones, I think I’ll be splitting it in smaller parts. And I’ll remove some years from the creation part as well.

They’re not always flat, but I’m having some troubles creating them, I can run the program for 12 hours and not getting any of them (with very wide parameters), so maybe is because my Data Range and my Data Parts, will take a look on it.