| 📉 Honest update: the previous version did not hold up in live trading. From 1 October to 8 November its own trades finished at −$178, and the live win rate fell to about 40%. On 10 November I replaced it with version 3 — rebuilt from the ground up, with a 75.6% win rate across 43 months of backtest and a fixed stop loss on every trade. |
In October I upgraded the EA three to four times, and it still was not enough. I could have adjusted a few more settings, but that only fits the past. So this time I rebuilt the whole thing. Here is what went wrong, what changed, and how you can check v3 for yourself.
1️⃣ WHAT WENT WRONG
The old version looked solid in every backtest. Live, it won far less often than the tests said it would.
| Old version · live, 1 Oct – 8 Nov 2026 | 🔴 Result |
|---|---|
| EA trades only | −$178 |
| Live win rate | about 40% |
When live trading keeps drifting this far from the tests, small fixes do not help. The strategy itself had to change.
2️⃣ WHAT VERSION 3 IS BUILT AROUND
| 🎯 | Two goals at the same time: a higher win rate and lower drawdown. Every part was re-tested, and only the parts that improved both were kept. |
| 🛑 | A fixed stop loss on every position. No grid, no martingale, no loss averaging. Lot size never grows after a loss. |
| 🧮 | Margin control. Free margin is checked before every order, and total used margin stays within 30% of equity. |
| ⚙️ | Simple to run. One XAUUSD M5 chart. The only real choice is your lot size. |
3️⃣ WHAT THE BACKTEST SHOWS
Version 3.06 · March 2023 to November 2026 (about 43 months, rising and falling gold) · every tick with random delay · 99% history quality · about 3,290 trades.
| Setting | Net profit | PF | Win rate | Relative equity DD |
|---|---|---|---|---|
| Auto Lot · $1,000 | +$254,094 | 2.27 | 75.6% | 17.5% |
| Fixed x1 · $1,000 | +$15,313 | 2.14 | 75.6% | 14.0% |
| Fixed x5 · $10,000 | +$76,784 | 2.14 | 75.5% | 7.4% |
The win rate is close to double what the old version managed live. For larger accounts, Fixed Lot at 0.05 lot per $10,000 kept relative drawdown at 7.4%. The full reports are on the product page.
4️⃣ WHY I AM NOT ASKING YOU TO TRUST THE BACKTEST
| ✅ The old version also looked good in testing. • That is why a backtest is where the work starts, not where it ends. • v3 has been trading live since 10 November on the same $1,000 IC Markets account. • Earlier records on the signal (max drawdown 26.7%, 5 manual trades) belong to the old version. Please judge v3 from 10 November onward. I would rather rebuild than sell something I do not trust. |
| ⚠️ The figures above are historical simulations, not live results, and they are not a promise about the future. Every GoldDragon trade has a fixed stop loss, and that is why I publish the live results openly. |
📅 EVERY TWO WEEKS, GOOD OR BAD
Every two weeks I check the live account against a backtest of the same period and post a Backtest vs Live comparison in the product comments — same period, same conditions, good results and bad ones alike. First v3 update: around 24 November 2026.
🔄 ALREADY USING GOLDDRAGON SAFE?
Update from the Market tab in your terminal, then check your lot settings. Auto Lot now uses x1 per $1,000 of balance (up to x20), and Fixed Lot goes up to x20. For lower risk, use 0.05 lot per $10,000.
5️⃣ WANT TO RUN V3 YOURSELF?
| 🐉 Launch price for the first 5 buyers: $397. The price then rises in stages to a final $1,297. Your price never changes after purchase, and all updates are free. 🔑 Not sure yet? Rent it from $59 and watch v3 trade on your own account — the same way you can watch it on the live signal. 👉 See GoldDragon Safe — buy or rent |
| 📡 | GoldDragon live signal — every trade published automatically |
| 🐉 | GoldDragon Safe — launch price $397 · rental from $59 |
| 📖 | Why the Profit Factor is around 2.0 — and how to spot martingale in any EA |
— Joon Jang (SAIYAN_ST)
