If you followed GoldDragon from launch, you may have noticed that the first live signal disappeared and a new one started in October. This post explains why. Short version: live trading showed me weak spots, I fixed them, and I restarted the track from zero rather than hide the old numbers behind new ones.
1. What went wrong with the first signal
The first signal started in September on a small account of about $253. I chose a small account on purpose, to prove the EA works with modest capital. In practice, that was a mistake:
- A small balance makes every loss look large in percentage terms. The drawdown on the signal page looked worse than the strategy really is.
- Watching the live trades next to the backtest, I could see that some modules were not behaving the way I wanted. Some losses were bigger than they needed to be.
A signal is a public promise. I did not want subscribers or buyers to judge GoldDragon on a version I already knew I was going to change. So I switched it off.
2. What I changed
GoldDragon runs 7 independent modules. Each one handles a different market condition, so when one struggles, others usually carry the account. After reviewing the live trades, 4 of the 7 modules got an upgrade. Entry ideas stayed the same. The changes were about how trades are managed and how losses are limited.
| Module | What changed | Result (module alone) |
|---|---|---|
| Module A | One position split into 4 smaller parts with staged take-profits. Stop moves closer, then to breakeven, as each part closes in profit. | PF 2.69 → 3.34 Largest single loss −$97 → −$47 |
| Module B | Now needs two independent signals to agree on direction before entering. Fewer trades against the trend. | PF 2.38 → 3.07 Drawdown 32% → 25% |
| Module C | Split into 2 parts with a breakeven step after the first target is hit. | Smaller losses on trades that turn back |
| Module D | Stops trading much earlier when the market is choppy and keeps going against it. | Drawdown 12.9% → 7.8% Worst losing run −$116 → −$68 |
Then a few more rounds
The first round was the big one, but I didn't stop there. Over the following weeks I released several smaller updates:
- Tighter hard stop-losses on several modules, so the emergency exit sits closer to where a trade is clearly wrong.
- A new first partial exit on one module, taking part of the profit earlier.
- One module's entry redesigned from scratch, with a simpler trigger and fixed take-profit and stop-loss.
- A small bug fix in how partial closes are calculated. A rounding error could close a slightly wrong lot size.
Every update is free and is delivered automatically through the Market, now and in the future.
I'm not publishing the module names or exact rules. That's the part of the EA people pay for, and keeping it private protects anyone who runs it.
3. The combined result
The goal was not a prettier profit number. The goal was to stop the worst trades from hurting so much. Here's the full 7-module EA, first version against the current version, on the same test:
Same compounding settings, $1,000 start, XAUUSD M5, January 2025 to September 2026, real ticks, 99% history quality.
| First version | Current version | |
|---|---|---|
| Net profit | $63,845 | $68,857 |
| Profit Factor | 1.93 | 2.02 |
| Equity drawdown (relative) | 24.36% | 21.03% |
| Largest single losing trade | −$452 | −$259 |
| Largest losing streak | −$677 (12 trades) | −$990 (16 trades) |
| Total trades | 2,581 | 2,169 |
The biggest change: the largest single loss dropped by 43%. Profit Factor went up, drawdown came down a little, and the EA now skips about 400 trades it used to take.
One trade-off: because positions are now split into smaller parts, losing runs can last a little longer in choppy weeks. Each loss is smaller, so the account handles them better.
This is also why the Profit Factor moved from 1.9 to 2.0 between my earlier posts. It didn't come from re-optimizing the same rules on the same data. It came from changing how trades are managed.
4. Realistic expectations
- Live will be lower than the backtest. Spreads, slippage and execution always take something. Treat the backtest as an upper reference, not a forecast.
- Losing days will happen. The new signal's first day was negative. Every trade has a hard stop-loss, so a losing trade closes at a known size. That's the system working, not failing.
5. The new live signal
Lessons from the first signal, applied:
- Starts with $1,000, matching the recommended capital, so drawdown percentages are not distorted by a tiny balance.
- One signal for all three grades. Safe, Standard and Aggressive share the exact same entry and exit logic, so one track shows how the strategy really trades.
- Same rules as always: no grid, no martingale, fixed stop-loss on every trade.
📈 GoldDragon Live Signal (MQL5)
📊 The same account is also tracked on Myfxbook under "GoldDragon Live".
If you're thinking about buying, my honest advice is the same one I give in private messages: let the signal run for a month and look at it yourself. You can also download the free demo and run it in the Strategy Tester on your own broker's data. I'd rather you buy with realistic expectations than be disappointed later.
Related posts
- Why My Profit Factor Is Only 2.0 — And How to Spot a Martingale EA
- One Strategy, Three Grades: How to Choose the Right GoldDragon
- Compounding Is Not Martingale: The Same Strategy, Two Very Different Numbers
- Seven Strategies, One Account: No Martingale Needed
- 20 Years, One Rule: What I Learned From the Accounts I Blew Up
Product pages: GoldDragon Safe | GoldDragon Standard | GoldDragon Aggressive
Questions are welcome in the comments.
— Jun
