Chapter 4 [Middle Part]: Breaking away from cover bands and the trap of excessive optimization
IV. Release on a demo account and a life of only 3 days
“If I do this, I’ll never lose even if real money is at stake. But just in case, I’ll try releasing it to the actual market on a demo account.”
On Monday at the start of the week, I proudly set up my “Homemade Holy Grail EA” on a demo account and turned on the run button.
Tuesday morning.
When I checked MT4 on my smartphone, I was receiving steady profit notifications: plus 1,200 yen, plus 800 yen.
“Alright, it’s exactly like the backtest said. It’s functioning perfectly.”
However, the nightmare arrived on Wednesday night.
Right after the European market opened, an abrupt statement by a key official caused the market to form a trend all at once, leading to a sharp increase in volatility.
At that moment, my “Homemade Holy Grail” began to go completely haywire.
“Hey...! Why are you entering here?! It’s the opposite; you’re totally going against the trend!”
The program, which had been forcibly fitted to favorable, gentle year-long market data, failed completely the moment it faced the “unknown real market” whose wave properties had changed.
The finely tuned numbers turned into mere noise, triggering multiple stop-outs, and the unrealized losses snowballed in an instant.
And during Thursday’s daytime, in a cafe during sales break.
Beep, beep, beep—.
From the smartphone speaker came the cold, familiar forced stop-out sound.
【Demo account: Stop Out】
The “Holy Grail” that was only tuned to a stable one-year market collapsed in just three days when subjected to forward testing. It crashed in just three days.
V. Do not cater to a specific past period — toward long-term testing and structural defense —
Facing a demo screen filled with bright red stop-out histories, I held my head and froze.
“Why…? The data should have yielded perfect numbers…!”
In despair and confusion, I remembered a maxim from a veteran programmer who posted on an overseas developer forum.
《Do not cater to a specific market》
A curve that has had its parameters fiddled to fit only convenient periods is nothing but an “after-the-fact painting (curve fitting).”
What truly survives in the market is not a system that relies on clever numerical fitting, but a “structural mechanism” that endures long data including more than a decade of harsh markets and shock periods.
What I was doing was not building a sturdy ship that would never sink no matter what storms came, but merely crafting a fragile wooden frame that perfectly fit the gentle waves through numerical fitting.
The future market will not stay as calm as past convenient periods.
“Stop deceiving yourself with short-term numerical fitting to win.”
Completely sever the sweet temptation of curve fitting that flatters a narrow period.
The goal should be not a system that relies on clever numbers, but a “multi-layer defense structure” that can withstand long-term historical shock markets for more than 10 years.
I spent days discarding the parameter files I had adjusted and faced a fresh code in MetaEditor again.
From here, the true core of GWR — a search for multiple independent blades to survive all-weather markets — began in earnest.
Six. The “pretend Holy Grail” thrown into the trash
The night the demo account balance hit zero and the inorganic log of forced stop-out streamed in.
I quietly dragged the parameter file, which I had pieced together by fitting numbers to a convenient “one year,” into the desktop recycle bin over several days.
“Right-click, empty recycle bin.”
The moment I pressed “Yes” on the confirmation dialog, the sweet illusion of easily beating the market with superficial numerical fitting completely vanished from my chest.
The asset curve built by pandering to a past period would not shield me in the future market.
The future market will inevitably bring history’s abrupt changes like the COVID-19 shock, months-long one-way trends, and storms of unpredictable key official statements.
“Even if you expose it to more than 10 years of rough seas and all shock markets’ tick data, build a structural strength that never folds. Nothing else can be entrusted with your family’s future.”
As a daytime office worker sweating on sales and by night facing MetaEditor, my development philosophy has shifted 180 degrees—from “maximizing profits through numerical fitting” to “designing multi-layer defense capable of surviving all weather.”