An Expert Advisor (EA) showing a perfectly smooth equity curve and a 95% win rate: that is exactly what should worry you, not reassure you. Most trading robots sold online do not survive real market conditions — and many hide a risk of ruin behind attractive numbers.

Here is how to tell a healthy robot from a scam, in a few minutes and without being a quant.

The 4 signs of a genuinely reliable EA

1. Controlled drawdown. Drawdown (the worst decline from a peak) is the best judge of a robot's survival. Below 10%, it is healthy. Above 20-25%, the risk becomes serious. An EA can be "profitable" and still ruin you if its drawdown exceeds what you are able to absorb.

2. A credible profit factor. Profit factor = gross profit / gross loss. Above 1.5, it is solid. Be wary of a profit factor of 5 or 10: it is almost always the sign of a track record that is too short or over-optimised.

3. A sufficient number of trades. Thirty trades prove nothing. It takes several hundred trades over several months for a result to be statistically credible.

4. An equity curve that breathes. A real curve rises, falls back, then resumes. A perfectly straight 45° line is a red flag: it is either an over-optimised backtest, or a martingale that has not blown up yet.

The 4 red flags of a scam

1. A hidden martingale. The robot doubles (or triples) position size after each loss to "win it back". It works… until the day a losing streak empties the account. This is the number one cause of ruin.

2. No stop-loss. Positions left open without a safety net: a single adverse move can wipe out months of gains.

3. An unrealistically "perfect" backtest. A 99% win rate, not a single losing month, a curve without a blemish: nobody trades like that. The backtest was adjusted after the fact to look good (over-optimisation / curve-fitting).

4. No transparency about real drawdown. An honest vendor shows their worst-case scenario. A vendor who only talks about gains has something to hide.

Backtest ≠ live

One crucial point: a backtest is a simulation on past data, optimised to look good. The only real judge is live behaviour (also called "forward" performance). Plenty of robots that look magnificent in a backtest collapse the moment they meet the live market (real spreads, slippage, news). Always ask for a real track record, not just a backtest.

Check your robot in 60 seconds

Rather than calculating everything by hand, import your MT4/MT5 statement: Judgebot gives it an A/B/C/D grade, automatically detects martingale behaviour and computes 12 risk metrics — free of charge.

👉 Scan your EA for free and get its full diagnostic in under a minute.

In summary

  • Look at drawdown and profit factor first, not total profit.
  • Steer clear of martingales and curves that look "too perfect".
  • Insist on a real track record, not an optimised backtest.
  • When in doubt: scan before investing a single euro.