Week 1: “I’m Staying the Course.” What Happened Next Changed Everything

Week 1: “I’m Staying the Course.” What Happened Next Changed Everything

7 October 2026, 13:53
Joon Jang
0
103
📌 Looking back, 10 November 2026: in this post I wrote that I was staying the course. A few weeks later, I did not. The live account showed me something this backtest could not, and it led to a full rebuild: GoldDragon Safe v3. The original report is kept below as it was written, and what I learned is at the end.

GoldDragon · XAUUSD M5 · Honest progress report · 7 October 2026

It has been one week since I deployed the new version of the code. To be honest, this week has been a struggle.

But one week is far too short to judge this strategy. It should be evaluated over at least one full month. Below is what the backtest said about normal monthly behaviour, including the bad months.

Test conditions (version at the time): XAUUSD M5 · every tick based on real ticks · fixed lot · initial deposit $1,000 · no compounding · Jan 2025 to early Oct 2026 (~21 months, continuous run). Monthly figures are read from the tester's monthly chart and rounded.

 

1. Summary

2025 2026 (Jan–Sep) Overall (~21 months)
Net profit +$5,537 ≈ +$10,500 +$16,051
Average month ≈ +$460 ≈ +$1,170 ≈ +$760
Worst month ≈ −$80 (Aug) ≈ +$500 (Aug) ≈ −$80
Positive months 11 / 12 9 / 9 20 / 21
Profit factor 1.83 2.44 2.15
Max equity drawdown – – $527 (4.5%)
Longest losing streak – – 16 trades (≈ −$190)

 

2. Month by month (net, approx.)

Month 2025 2026
Jan +$280 +$2,400
Feb +$890 +$1,780
Mar +$370 +$580
Apr +$480 +$740
May +$820 +$660
Jun +$610 +$1,680
Jul +$300 +$1,560
Aug −$80 ◀ only losing month +$500
Sep +$340 +$520
Oct +$640 in progress
Nov +$500 –
Dec +$340 –

 

What this told me at the time: even with only one losing month in about 21 months of backtest, there were plenty of losing weeks and losing streaks of up to 16 trades. A weak first week looked like a normal part of this equity curve.

 

3. Why I planned to stay the course

  • No martingale, no grid. Every entry has a fixed stop loss, so each loss closes at a known size.
  • Changing settings in the middle of a drawdown is one of the most common ways traders turn a normal dip into a real loss.

Good weeks or bad weeks, I said I would keep sharing the results openly. That part I have kept.

 

4. Looking back: what the next month actually showed

✅ The rule in this post was right: judge an EA over a full month, not a week.

❗ But the month gave a clear answer. Over October the live win rate stayed far below the backtest. That was not a normal bad week inside a good curve. It meant this version did not hold up live.

🔄 So I did the thing I warned against, and I want to be honest about why. Staying the course is right when live trading matches the test. When it clearly does not, staying is just hoping. I stopped patching and rebuilt the strategy set around a higher win rate and lower drawdown.

 

5. The same rule now applies to v3

GoldDragon Safe v3 has been running on the live signal since 10 November 2026. Judge it the same way: over at least a month, not a week. The difference this time is that you can watch that month happen trade by trade, and every two weeks I post the live result next to a backtest of the same period in the product comments.

📡 GoldDragon Live — IC Markets, $1,000 start, v3 since 10 November 2026
📖 Why GoldDragon Safe moved to v3
🐉 GoldDragon Safe v3 — launch price $397 · rental from $59

Backtest results do not guarantee future performance. Live results may differ due to spread, slippage and broker conditions.

— Joon Jang (SAIYAN_ST)