A 75% Win Rate Usually Means Martingale. Here Is Why Mine Doesn’t.
| 📌 Updated 10 November 2026 for GoldDragon Safe v3. This post first used an earlier version, which won about 50% of its trades with winners about twice the size of its losers. v3 has a different shape: it wins far more often. Below I explain why that is still not martingale, and how you can check it yourself. |
When people see GoldDragon's backtest, the first question is often: "Where does that profit come from? Is it martingale?"
And now that v3 wins about three trades out of four, I expect that question even more. A high win rate is exactly what a martingale EA shows off. So here is the honest answer, with numbers you can check yourself.
1. Win rate alone proves nothing
A 75% win rate can come from two very different systems:
- A martingale or grid. It wins 90% or more of the time because it refuses to close losers. It averages down or doubles the lot until the market comes back. The losses are not gone. They are stored up, and one day they arrive all at once.
- A system with real stops. It wins often, every loss is closed by a fixed stop loss at a known size, and the next trade starts at the same size as before.
From the win rate alone you cannot tell them apart. From the numbers below, you can.
2. The numbers that rule out martingale
GoldDragon Safe v3.06, Fixed Lot x1 (no compounding), $1,000 start, XAUUSD M5, March 2023 to November 2026, every tick with random delay, 99% history quality.
| Win rate | 75.6% |
| Average win vs average loss | $11.53 vs $16.57 |
| Largest losing streak | 11 trades, −$94.79 |
| Profit factor | 2.14 |
| Trades | 3,296 (about 75 per month) |
| Relative equity drawdown | 13.97% |
| Recovery factor | 29.97 |
Here is what to look at:
🔍 The average loss is only about 1.4 times the average win. In a martingale, the rare loss is usually many times larger than a typical win, because it is the sum of everything that was held and doubled. Here, losses stay close in size to wins, because each one is cut by a fixed stop.
🔍 The worst losing streak is 11 trades for under $95 on a $1,000 account. A martingale cannot survive 11 losses in a row without its lot size exploding. GoldDragon simply takes 11 normal-sized losses and moves on.
🔍 A recovery factor of about 30 means total profit was roughly 30 times the deepest drawdown. The profit is not one lucky run: summed over the test, every weekday from Monday to Friday ended in profit.
3. The test that settles it: change the lot size
Run the same test at Fixed x5 on $10,000 and Fixed x10 on $20,000. Profit goes from $76,784 to $153,325, almost exactly double, and the profit factor (2.14) and win rate (about 75.5%) do not move.
A martingale cannot do that, because its position size depends on the last loss, not on your setting. The free demo lets you run this check in the Strategy Tester in a few minutes.
4. What about compounding?
With Auto Lot, GoldDragon grows the lot as the balance grows: x1 for every $1,000, up to x20. That is compounding, and it is the opposite of martingale. Lots grow after the account has gained, never after it loses. The full fixed-lot versus compounding comparison is here:
👉 Compounding Is Not Martingale: The Same Strategy, Two Very Different Numbers
5. See it live
Backtests are an upper reference. Live results will be lower because of real spreads and slippage. And a different shape from the last version means one thing above all: v3 has to prove itself live. That is why every trade is public, and why every two weeks I post the live result next to a backtest of the same period in the product comments.
| 📈 GoldDragon Live — $1,000 start, running v3 since 10 November 2026, every trade published automatically 🐉 GoldDragon Safe v3 — launch price $397 · rental from $59 · free demo No grid. No martingale. A fixed stop loss on every trade. 📖 Want to dig deeper? Why GoldDragon Safe Moved to v3 · How to Spot a Martingale EA · Why I Switched Off My First Signal · 20 Years, One Rule |
Backtest figures are historical simulations and do not guarantee future performance.
Questions are welcome in the comments.
— Joon Jang (SAIYAN_ST)


