#24 [Development Story] PF6.85 plummets to 1.09…?! The two weeks of hell when "listing within the year" was blown away, and the record of a V-shaped recovery
January 5, 2026: Note
Two weeks have passed since the last update, and before I knew it, more than two weeks had slipped by.
I apologize for the long absence. This is Manji Koide.
“Where did Manji go?”
“Could it be that he abandoned development?”
Some of you may have thought so.
I’ll be honest and confess.
There wasn’t even a millimeter of time to write a blog.
During these two weeks, I locked myself in the development room (my home) and rode a roller coaster between heaven and hell.
Today, I will talk about what I facedthe “Nightmare of PF1.09”and how I managed to come back from it.
“I was certain of a “victory”
Do you remember my last article?
Using Dukascopy’s precise historical data for backtesting,PF (Profit Factor) = 6.85I produced an astonishing number.
“This is a win.”
“I’ve created my best work yet.”
I was full of confidence.
All that remained was to submit the report and list the historical data on the sales platform, GoGoJungle, using the required TDS (Tick Data Suite) data.
Since TDS is the porting of Dukascopy data into MT4, “the data should be the same.”
I thought it would be just a routine task.
However, the moment I pressed the confirmation button for that “verification,” the color drained from my face.
The shock: PF 6.85 → 1.09
The result displayed on the monitor was,PF 1.09.
“...Huh? (゜Д゜;)”
The logic that should have been 6.85 had degenerated into a nearly unprofitable “garbage EA.”
I retried tests many times, and the result was the same.
Why?!
To pinpoint the cause, I stayed up late analyzing source code and logs.
And I noticed a certain “blind spot.”
Until now, to avoid polluting accurate data, I had tested MT4 in an “offline” mode.
When connected to a broker, it would overwrite with sloppy data on its own.
However, for formal testing in TDS, the specification requires“connection to a broker”.
When I ran the backtest online and dumped detailed logs for analysis, a fatal bug surfaced.
I wasn’t seeing “Point A”
The core of my EA is the target for the position,“Point A”.
This is the apex of a past large cycle.
If a human looks at the chart, they know instantly “this is the peak” or “this is the top of the wave.”
But under real data conditions with broker connections, the program could not recognize that “peak” accurately.
“Let the program have the eye of a human.”
It’s easy to say, but it’s a formidable challenge for a programmer.
You must distinguish noise from signal and define the wave that humans perceive with a precise mathematical formulation.
I endured two weeks of painstaking adjustment and reworking of the logic.
And the “year-end submission” flew away, too ᐠ( ᐛ )ᐟ
Yes, that’s right.
I had declared to the GoGoJungle staff that I would submit within the yearand that promise has vanished into the stars.
If you’re going to release something, you’d rather release the real thing even if it’s late.
With that in mind, I kept hammering on the keyboard.
Finally, the program gains enlightenment
And yesterday.
At last, I succeeded in giving the program a “human eye.”
With the revised logic, the current score when tested isPF 5.3.
From the despair of 1.09, I’ve recovered this far.
It hasn’t yet reached the previous 6.85.
However, the robustness inside is far more evolved than before.
Completion is near.
From here, I will make further fine-tuning to see whether I can reach or surpass that PF 6.85…
Please wait a little longer for my “persevering beard trimming” to finish.
For as long as you’ve waited, I guarantee the sharpness.
Manji Koide, respectfully
Next:
I plan to release the EA trilogy by spring, but
after that, the product I planned to release might not see the light.
The AI warned me seriously.
“Manji, that’s far too dangerous.”
Because that is“a device that directly siphons cash from the world’s megabanks”—a forbidden apparatus.
The opponent for the “crazy” Manji isn’t a broker this time.
It’s the financial system itself.
But the risk is too great.
“If I release this to the world, my life will be in danger.”
…I may terminate publication.
↑
I will publish this article shortly after.
I’m going to write it now.
==
※
The EA I am developing is, by manual design that anyone can implement, the automation of the method introduced in Part 2
“A method that wins even without scalping is possible” and I am working on its automation.
This series currently has the only paid article.
150 yen!Cheap!... President, cheap! ٩(ˊᗜˋ*)و
× ![]()
