#23 [Development Secrets] A "heated" call from GoGoJyan, and the "harsh reality" faced during TDS verification…! The reason I am still burning with passion
December 20, 2025: Note
A “Warm Cheer” from Gogojan
In the past few days, I received back-to-back calls from Gogojan for two days in a row.
They spoke in a polite tone, but if you translate their request...
“Please list it soon!”
“We’re looking forward to your listing within the year!”
...That was a very heartfelt cheer.
(Note: In my head it naturally sounded like, “Hurry up and publish it, you! We’re counting on you this year!” but this is just a reflection of their expectations. Haha)
Among that, one important “regulation” was mentioned.
“Please use history data from Tick Data Suite (TDS) for the test results.”
I said.
“No, no, I’m using data from Dukascopy. TDS is just a tool to download data from Dukascopy to MT4, so the data should be the same, right?”
Then Gogojan’s answer was clear.
“That is our regulation, so please use TDS data (we won’t budge on this, please handle it properly!).”
...Yes, understood
When in Rome, do as the Romans do.
As instructed, I downloaded the historical data from TDS using the official procedure, and decided to“do it myself”and run the test again.
“The baptism of TDS and the shattered dream of PF6.85”
With high spirits, I tried to run the Strategy Tester with the “Tick Data” setting, but MT4 issued a merciless warning.
“Log in to your broker account.”
I was taken aback by this.
If I log in to the account, the broker’s server may deliver “sloppy data (rough data),” which could overwrite the carefully prepared historical data.
That’s why I had been testing with offline charts up to now.
====
But I had no choice but to proceed.
I was confident.
Because in previous tests
“Spread = 28 (2.8 pips) fixed spread with a profit factor (PF) of 6.85”
I had achieved an astonishing number.
“If we can win with a fixed 2.8 pips spread, in broker’s variable spreads there would be even narrower moments and even more extreme numbers!”
With that expectation, I pressed Start.
====
And when I saw the results, I was stunned.
...Total defeat.
My EA, which had PF 6.85, when run in the TDS environment,PF 1.09(cry) became that value.
====
I would never engage in dishonest tricks like “curve fitting (over-optimization)” to fit past data.
So I tried adjusting parameters and re-challenging, but... the result didn’t change.
Why is the result so different?
The reason the EA that was strongest with fixed spreads performed poorly in a realistic variable-spread (TDS) environment.
A calm analysis reveals the following realities.
1. The reality of “real spreads”
I found that the spread widens more than I expected exactly when I want to enter (when the market moves a lot).
2. The silent assassin “negative swap”
This was an unexpected blind spot.
My EA is designed to sell the USD/JPY (short).
Carrying a USD-selling position in the current market meansyou pay a high interest rate (swap points) every dayas a cost.
I thought, “I won’t hold for long, so it’s okay.”
However, TDS’s accurate data is cruel.
While waiting for the market to go down for a few days,“stay cost (swap payments)”acted like a savings of damage from a low kick, eroding profits.
Even when winning, profits shrink; if you lose, you get hammered by “stop-loss + swap”… this was the real cause of PF collapse.
3. The shape of deception
Tempestuous noises and wicks, unseen in clean chart data, were confounding me.
4. The program’s “vision” deficit
This is the main reason (explained later).
“Human Eyes” VS “Program Eyes”
－ Capture Point A
The hair-removal EA I’m developing uses
“to scalp the top of the previous big wave”
as its unique approach.
What matters most here is
“the previous大 wave’s peak (Point A)”
to capture accurately.
To a human, this is incredibly easy.
You can tell in 0.1 seconds, “Look, that peak is the highest.”
If a human trades by this method, it’s easy to win.
I won’t lose at first (for details,see Episode 2).
Most EA in the world boast about “fast processing that only machines can do,” but
my EAaims to have a living human’s discretionary trading done by the program as is.
Why didn’t I win in this test when humans can?
Because,the program cannot recognize Point A like a human eye.
The program cannot “see the mountain’s shape.”
It can only judge by “13 days ago” or by numbers.
How can I give the program a human perspective to accurately capture Point A?
The task is highly challenging, but I see a glimmer of a solution.
It’s extremely difficult, but for some reason I’m finding it incredibly enjoyable.
Afterword
This time I realized something.
I’m glad Gogojan is strict...
( ˊᵕˋ ; )
If I had listed based only on that fixed-spread test result...
Gogojan requires the source code at listing.
Furthermore, they perform strict tests before listing, so any fake results would be uncovered immediately.
All EAs currently listed on Gogojan have passed this test and are genuine.
Unlike places flooded with fraudulent products like the old scam sites, this is a place with strict referees.
It’s a great arena for my new challenge.
This is the perfect arena for me (smiles).
Next:
Has the program gained human eyes?
※
The EA I’m developing is intended to automate the approach shown in Episode 2, which anyone can implement manually.
“Even without scalping, it’s possible to win big”— the automation of that method.
This is currently the only paid article in this series.
150 yen!So cheap!... President, so cheap! ٩(ˊᗜˋ*)و
× ![]()
