How to spot a fake EA before it spots your account.

A bad Expert Advisor rarely looks bad. It looks perfect: a smooth curve, a 95% win rate, hundreds of happy reviews. Here are the red flags that give it away, and a five-step test you can run before paying for any EA, including mine.

Marketplaces like the MQL5 Market are full of genuinely good work by honest developers. They are also full of products built to sell rather than to trade. The difference is rarely visible in the headline numbers, because the headline numbers are exactly what a bad EA is designed to get right.

The good news: fake and fragile EAs leave fingerprints. Once you know where to look, most of them give themselves away in ten minutes. This guide is written so you can check any product, on any marketplace, without writing a line of code.

The one rule

An honest seller shows you how the strategy loses. A dishonest one only shows you how it wins. Every red flag below is a variation of that.


Why bad EAs look so good

Selling an EA pays the same whether it works or not. That creates a simple incentive: build whatever produces the most attractive screenshot. And the most attractive screenshot in trading comes from strategies that hide their losses rather than avoid them.

The classic example is the martingale or grid EA. When a trade goes against it, it doesn't take the loss. It opens another trade, often bigger, and waits for price to come back. Most of the time price does come back, and the whole basket closes in small profit. For months the results look flawless. Then one day price doesn't come back, and the account is gone in a week.

Balance curve versus equity curve of a grid EA The balance line climbs smoothly while the equity line repeatedly falls far below it, then collapses at the end. balance equity the one bad week →
Illustration, not real data. A grid or martingale EA keeps losing trades open, so the balance only records the small wins. The equity shows the losses the balance is hiding, until the day they are finally closed.

The same trick works in reverse for sellers. A strategy that makes 2% a month with a 15% drawdown is honest and hard to sell. One that shows 30% a month with a "3% drawdown" sells itself, and nobody asks which curve that 3% was measured on.


Red flags in the pitch

Start with what the seller says, before you look at a single number.

⚠
Returns that sound like a salary Consistent double-digit monthly returns are not a trading edge, they are a risk profile. Anything that makes 20% a month can lose 20% a month, and usually more.
⚠
"No losing months" or "never loses a trade" Every real strategy has losing streaks. Either the losses are being hidden inside open positions, or the track record is too short to have met a bad market yet.
⚠
Buzzwords instead of an explanation "AI", "neural network", "quantum" and "smart recovery" are not strategies. A good seller can tell you in two sentences what the EA waits for, where the stop goes and why it exits. If they can't, or won't, assume you wouldn't like the answer.
⚠
Pressure instead of proof "Price goes up after 10 more copies." Rising prices are common and not dishonest by themselves, but urgency is never evidence. If the only reason to buy today is the countdown, wait.

Red flags in the numbers

⚠
Only a balance curve, never an equity curve This is the big one. The balance only moves when trades close. The equity includes the losses still sitting in open trades. A grid EA can show a ruler-straight balance while its equity is 40% underwater. Always ask for equity, or look for it in the live signal.
⚠
A very high win rate with no stop loss A 90%+ win rate combined with no hard stop loss, or a stop ten times wider than the target, means many small wins paid for by rare, enormous losses. Check the trade list: if losing trades are few but huge, you've found it.
⚠
Lot sizes that grow after losses Open the backtest's trade list and look at the volume column. If the lot size doubles or climbs after a losing trade, that is martingale, whatever the description calls it.
⚠
A backtest that starts suspiciously late Why does the test begin in March 2023 when data exists back to 2015? Often because the earlier years were bad. Ask for the full history, or run it yourself.
⚠
Perfect results on the exact period it was tuned on A strategy optimised on 2019–2024 and then shown on 2019–2024 has memorised that history. That is overfitting. Look for results on out-of-sample data the EA never saw during development.
⚠
Low-quality backtest data MetaTrader 5 reports a History Quality percentage. Well below 90% means much of the price movement was simulated rather than real, which flatters strategies with tight stops. My backtest guide explains what good looks like.
⚠
Live results that are short, demo-only or missing Three months of live trading proves very little; grids survive that easily. A live signal on a demo account proves even less. Look for a real-money account, running for six months at least and ideally over a year, that has lived through at least one ugly market.

Red flags in the seller

⚠
No person behind the product No name, no history, no way to ask a question. You are trusting your money to the software; it's fair to know who wrote it.
⚠
Old products that quietly disappeared Look at the seller's profile. Many products, a few recent ones and older ones renamed or removed can mean strategies that blew up and were replaced with fresh listings.
⚠
Reviews that all arrive at once A burst of five-star reviews in the first days, all short, all praising "support" rather than results, deserves a closer look. The most useful reviews are older ones written after months of use, including the three- and four-star ones.
⚠
Evasive answers about drawdown Ask directly: "What was the worst drawdown, and how long did it take to recover?" A developer who knows their system answers that immediately, with a number. Vague replies about "proper risk management" are an answer too.

The five-minute live-signal check

If a product links to a live signal on the MQL5 Signals service, Myfxbook or a similar tracker, it is the most valuable page you can read. Look at four things:

  • Real or demo. Only a real-money account counts as a track record.
  • Age. Months, not weeks, and ideally through a market that hurt other strategies.
  • Equity drawdown. The deepest fall of equity, not balance. If it isn't shown, that is information.
  • How much of the account is in use. On MQL5 Signals this appears as deposit load; elsewhere look at margin used. Constantly high values mean positions far too large for the account, the typical footprint of grids.

Test before you pay: five steps

You don't have to take anyone's word for anything, mine included. Products on the MQL5 Market can be downloaded for free as a demo that runs in the Strategy Tester. Use that.

  1. Read the description for three answers: what the EA waits for, where the stop loss goes, and how it exits. If any is missing, ask the seller before going further.
  2. Backtest it yourself, on your broker's data. Run a quick first pass on 1 minute OHLC, then confirm on Every tick based on real ticks and compare. A big difference between the two is a warning on its own. See modelling mode.
  3. Read the trade list, not just the report. Check the volume column for growing lots, and the losing trades for size. Then find the worst stretch of the test and ask yourself honestly whether you would have kept the EA running through it.
  4. Size it before you run it. Work out what the stop distance means in money at your risk level with the lot size calculator. If the smallest lot is already too much risk for your account, that's your answer.
  5. Demo first, then small and live. A few weeks on a demo account, then real money at a fraction of the size you intend to use. Scale only once live behaviour matches what the backtest led you to expect.

The fake-EA checklist

  • The seller explains the strategy, the stop loss and the exit in plain words
  • Every trade has a hard stop loss, and it isn't many times larger than the target
  • Lot sizes don't grow after losing trades
  • An equity curve is shown, not only a balance curve
  • The backtest covers several years and was not only run on the period it was tuned on
  • History quality is high, and real-ticks results are close to the quick test
  • A real-money live signal has been running for six months or more
  • The seller answers "what was the worst drawdown?" with a number
  • You backtested it yourself on your broker's data
  • You ran it on demo, then live at small size, before scaling

Hold me to the same standard

I sell EAs, so it would be strange to write this and ask you to skip the checks for mine. Please don't. Here is where to find everything this guide tells you to look for:

  • The strategy in plain words: opening-range and session breakouts, explained step by step in the roadmap.
  • Hard stops, no martingale, no grids: every trade has its stop loss from the start, and lot sizes never grow to recover a loss.
  • Full backtest reports: published for each trading system, with the trade statistics and history quality.
  • Live, verified results: my real account is tracked independently by FX Monitor, drawdowns included, and the charts are streamed on YouTube. See the live section and the portfolio dashboard.
  • A person you can ask: I answer questions before and after you buy at [email protected].

And if you would rather start without paying anything, the free EAs are the easiest way to put this whole checklist into practice.

Important Disclaimer

Past performance does not guarantee future results. Passing every check in this guide makes an EA more trustworthy, not risk-free. Any strategy can lose money, and all trading should be done with capital you can afford to lose.