Astraeus MT5
- Experts
-
Minh Truong Pham
Hi, I'm Pham, a programmer and trader from Vietnam. I write indicators and Expert Advisors for MetaTrader 4 and 5. I write every product myself from scratch, without borrowing third-party code.
How I build and sell EAs - 버전: 1.0
- 활성화: 10
$1,000 became $5,735 in a backtest on real XAUUSD tick data, 16 March 2020 to 22 October 2026, at 0.01 lot per trade with commission taken off. That's +$4,734.56 across 3,069 trades, profit factor 1.42, and all seven calendar years finished green.
Astraeus is a gold breakout EA for MetaTrader 5. Stop loss on every trade. No grid, no martingale, no averaging down.
It's a backtest, not a promise. The rest of this page is the detail behind those numbers, including the parts that aren't pretty.
The backtest, in numbers
Real XAUUSD ticks from a raw-spread account, 2020-03-16 to 2026-10-22, $1,000 start, fixed 0.01 lot per trade, $4 per lot commission taken off every trade, swap not included.
| Net profit | +$4,734.56 (+473%) |
| Profit factor | 1.42 |
| Trades | 3,069 (1,632 buy / 1,437 sell) |
| Win rate | 40.4% |
| Average win / average loss | +$12.91 / -$6.17 |
| Max drawdown (balance) | $414 |
| Positive months | 46 of 80 |
| Biggest single loss | -$32.43 |
| Most trades open at once | 5 |
Year by year, same 0.01 lot:
| Year | Result |
|---|---|
| 2020 (from 16 March) | +$409 |
| 2021 | +$397 |
| 2022 | +$46 |
| 2023 | +$815 |
| 2024 | +$426 |
| 2025 | +$1,871 |
| 2026 (to 22 October) | +$771 |
Since 2023 alone it's +$3,882 on 1,724 trades with a profit factor of 1.52, if you want the more recent picture.
What it costs next to what it made
Astraeus costs $299. At the default 0.01 lot, the 2025 backtest year made +$1,871, and 2026 had made +$771 by 22 October. That's 0.01 lot, the smallest size most brokers offer, on a $1,000 test account.
I'm putting those two numbers side by side so you can do the sum yourself. I'm not telling you 2027 will look like 2025. Nobody can, and 2022 made only $46.
How it makes money
Read the win rate again: 40%. Astraeus loses more trades than it wins, and that's how it's built.
Losers are cut at a stop that's sized to the price of gold, so they stay small. Winners don't have a fixed target. The stop trails behind them and the trade runs until the market takes it out, which is why the average win is about twice the average loss and why a few trades a year make $50 to $100 each at 0.01 lot.
Several models run on the same chart. They look for the same kind of breakout at different scales, from short moves to ones that took days to build, and each model holds one trade at most. The slower models trade less often but carry more of the profit: in this test the slowest one made +$1,308 and the fastest +$495. Every model was positive on its own.
I'm not publishing the entry rules. That's the part that took the time.
What trading it feels like
Honestly, it can be dull and then annoying. 34 of the 80 months lost money. The worst month in the test was October 2026 at -$267, and the deepest balance drawdown was $414, between May and August 2026. So the worst stretch is the most recent one, and you should know that before you buy.
You'll see runs of small stop-outs. Then one trade runs for hours and pays for the lot of them. If you switch it off after ten losers in a row you'll miss exactly the trade it exists for.
Safety
- Every order is sent with its stop loss attached. There's never a trade without one.
- The stop only moves in the trade's favour.
- One trade per model, five at most, all the same size. It doesn't add to losers.
- Before each order it checks the lot against your broker's limits and your free margin, trims the lot or skips the trade if the account can't carry it.
- No DLL, no outside connections.
Requirements
- Symbol: XAUUSD, or whatever your broker calls gold. It's built and tested for gold only. It will start on other symbols (every distance it uses is sized from price), but I haven't tested it there and I wouldn't trade it there.
- Timeframe: attach it to M1. It reads the bars it needs by itself, so the chart timeframe doesn't change the trades.
- Account: hedging. Two models can hold opposite trades for a while. On a netting account it keeps to one position at a time, which means fewer trades than the test.
- Spread: a raw-spread or ECN gold account is what the numbers above come from. On a wide-spread account expect less, because every trade pays the spread and there are around 470 trades a year.
- Capital: I'd start at $2,000 per 0.01 lot. The test ran on $1,000, but five trades can be open together and I'd rather have room to spare.
- VPS: recommended. It checks prices every minute, around the clock.
Inputs
| Input | Default | What it does |
|---|---|---|
| Fixed lot per trade | 0.01 | Size of every trade when Risk % is 0 |
| Risk % of balance per trade | 0 | Sizes trades from your balance instead of a fixed lot (0 = off). All models get the same lot |
| Magic number | 2026150 | Keeps its trades apart from other EAs on the account |
| Max spread at entry, points | 0 | Skips an entry while the spread is wider than this (0 = off) |
| Broker server time zone | Auto | Leave on Auto for live trading |
| Server GMT offset, hours | 0 | Only used in Fixed mode and in the Strategy Tester |
Everything else is fixed at the settings the numbers above belong to. There's nothing to optimise, and that's on purpose.
How to backtest it
Strategy Tester, XAUUSD, M1, model Every tick based on real ticks. Anything coarser gets the exits wrong, because it manages open trades minute by minute.
One setting matters in the tester: the time zone. Auto can't read the real clock there. If your broker's server runs on GMT+2/+3 with US daylight saving (most do), choose that option. If it runs on a fixed offset, choose Fixed and type the offset.
Your result won't match mine to the dollar. The table above comes from my own tick-by-tick engine, which I checked trade by trade against the MT5 Strategy Tester: January to October 2026, 353 trades against 344, 98% matched within two minutes. Different brokers have different ticks and spreads, so test it on yours.
Questions people ask
Can I turn some models off? No. They work as one set and that's what was tested.
Does it trade news? It doesn't read a news calendar. If your broker's spread blows out around news, set the max spread input.
Why no take profit? Because the big trades are where the money is. A fixed target would cut exactly those.
Can I raise the lot? You can, and everything scales with it, the drawdown too. Ten times the lot is ten times that $414.
Updates and support
I fix bugs and keep it working with new MetaTrader builds. Updates are included. If something looks wrong, write in the Comments tab with your broker, symbol name and a screenshot of the Experts log, and I'll look at it.
Past results don't guarantee future results. Trading gold with leverage carries risk. Start on a demo, then go small.
