1 Topic by dusktrader (edited by dusktrader 2025-10-06 02:01:34)

  • From: United States
  • Registered: 2012-02-06
  • Posts: 125

Topic: How is "Is strategy better?" calculated on Walk Forward

I was wondering if you could shed some light on how the system determines if a Walk Forward strategy is better or not?

When I performed Walk Forward on this one, it looks great to me on the surface. Only 1 segment was less profitable, and overall it achieved the same net profit as the original.

But the statistics seem pretty negative and they don't seem to reflect the graph. Is this a mistake?

[img]walkforward.png[/img]

[img]walkforward stats.png[/img]

Thank you

EDIT: sorry not sure how to attach images inline

Post's attachments

walkforward stats.png 89.91 kb, file has never been downloaded. 

walkforward.png 70.58 kb, 1 downloads since 2025-10-06 

You don't have the permssions to download the attachments of this post.
  • Registered: 2020-07-11
  • Posts: 27

Re: How is "Is strategy better?" calculated on Walk Forward

Hello

That’s not a mistake EA Studio follows a strict logic when deciding whether a Walk-Forward strategy is “better.”
It doesn’t judge by how the graph looks; it compares the optimized statistics against the original strategy’s statistics.

EA Studio checks key metrics like:

Net profit

Profit factor

Drawdown

Return/Drawdown ratio

Sharpe ratio

Balance line stability (R²)

If most of these values get worse after optimization, EA Studio automatically flags:

❌ “Is strategy better = No.”

In your case, all five out-of-sample segments ended profitable — that’s positive —
but the overall behavior is statistically weak and unstable.
The optimized version clearly degraded:

Profit Factor 1.59 → 1.20

Sharpe Ratio 0.18 → 0.07

Return/Drawdown 1.51 → 0.54

R² / Balance Stability 67 → 28

Net profit dropped, stagnation increased

So the system correctly marks “No.”
Even though every segment made money, the optimization didn’t strengthen the system it reduced robustness and efficiency.

Also, this strategy shouldn’t really be pushed through Walk-Forward Optimization (WFO) at all.
WFO isn’t meant to fix weak systems; it’s meant to validate and adapt already solid ones across different market periods.
With only around 45 trades and soft baseline stats (PF 1.59, Sharpe 0.18),
the sample size is too small to draw reliable conclusions — the test just amplifies noise instead of proving robustness.

  • Registered: 2020-07-11
  • Posts: 27

Re: How is "Is strategy better?" calculated on Walk Forward

Let me be real with you, man.
I read some of your old posts from 2012.
Thirteen years later, you’re still talking about the same things: smooth balance lines, 45° curves, “good-looking” charts.
But that’s not growth. That’s looping.

The strategy you just pushed through Walk-Forward Optimization wasn’t smooth at all it was noise.

That’s the problem.
You’re chasing the idea of a perfect line, but you’re not recognizing when the data is clearly broken.
A system with 45 trades and an unstable curve doesn’t need optimization it needs deletion.

WFO can’t fix that.
It doesn’t turn weak systems into strong ones; it just exposes what’s already fragile.

Thirteen years in the generator doesn’t make you experienced.
It makes you comfortable with illusion.
And comfort kills more traders than drawdown ever will.

Real growth starts when you stop trying to make a bad curve look smart and start asking why your so-called “smooth” ideas keep collapsing in stats.
That’s the moment you stop being a button-clicker and finally start thinking like a quant.

At some point, you’ve got to stop chasing perfect lines and start chasing truth even when it means deleting everything you built.
Because that’s the moment you stop being a user of tools and start becoming a builder of systems.

It’s clear that nobody has ever told you the truth on all your posts in 13 years, so I hope this is an a-ha moment for you.
You really need to go deep into metrics and robustness.
And one crucial point: your data must be long enough with 45 trades you can never know if a system is robust.

I’m sure you’ve had some lucky good systems over the years after 13 years that’s bound to happen but honestly, it’s painful to see someone post today what you just posted.
It tells me everything.
I hope you take this to heart.
It may sound harsh, but you need to wake up, because right now you’re living in an illusion about robots and how to build them.

  • From: United States
  • Registered: 2012-02-06
  • Posts: 125

Re: How is "Is strategy better?" calculated on Walk Forward

Hello Jurgen - thank you for your honest input here!

It's true I landed back on Mr. Popov's software in a weird serendipitous circle. I have been working on trading for a long time as you noted, but still no major traction that I'm happy with. I think the last time I took a serious look at Mr. Popov's software ... as I recall he was a sea captain (stood out to me as an interesting life career) and I think the software must have been only in alpha-stage development as I cannot find where I purchased anything related to FSB Pro. In a later cycle of algo study, I ran into Mr. Petko and at some point I landed back on Mr. Popov's creations again!

But I go through long phases of discretionary manual and then phases of algo trading. So all those years have not been focused on algo trading. But I recently shifted back to algo trading so it might take me awhile to realign everything in my mind. I don't know why my brain does this cycling, but I think it is from pushing too hard -- when I get burned out on one area of study, I shift gears. But I save the experience and learned nuggets. I can probably tell you all the things I found that DON'T work haha!

As an engineer and software developer, I have technical skills and I've tried to apply them progressively over the years. So my current stage will be figuring out how to properly assess algo strategies in general. From an infrastructure standpoint, I think I have everything mastered so at this point I'm just trying to learn how to read these statistics.

(I haven't read through any of my super old posts so I have no idea what I was bantering on about back then haha! Only know that I'm still here and kicking.)

Thank you very much for the detailed explanation about my original question regarding the WFO analysis.

I recently learned how to use the Reactor feature along with many optimization filter steps. So now my process involves focusing on these 3 pairs for correlation balance that I learned from Mr. Petko (EURUSD, GBPUSD, EURGBP) and I'm generating on the M15. This is just my new starting point so I'm sure it will evolve over time.

Adding Reactor has really helped pair down the universe of generated strategies. Now I'm just trying to get a deeper understanding of what the statistics are telling me.

My plan is to follow what I learned from Mr. Petko -- to generate constantly on pairs I'm looking at, using the actual broker data when possible, and then preparing a fleet of bots to trade in both demo and live accounts. Per Mr. Petko's teaching, I am currently working to find good strategies daily and after monitoring the demo account performance, I plan to identify the cream-of-crop Top5 and trade those in the live accounts.

Btw, nice to meet you!