Search for "best forex robot" and you will get dozens of rankings. They all share one trait: none of them can be verified. The figures come from the vendor, the tests are almost never reproducible, and most of these pages earn a commission on the sale of whichever robot they place at number one.
So this guide is not another ranking. It is a method you can apply yourself, to any Expert Advisor, to decide whether it deserves your capital.
Why the "best EA" does not exist
A robot has no absolute performance. Its performance is conditional, and it depends on four parameters only you know.
Your capital. A robot whose historical drawdown reaches 25% is workable on an account you can leave running for two years. It is unmanageable on capital you will need within six months.
Your broker. A scalper that takes 4 pips per trade loses its entire edge at a broker whose average spread is 1.8 pips instead of 0.6. The same robot, two brokers, two opposite outcomes, without a single line of code changing.
Your tolerance for inactivity. Some strategies only open a handful of positions per month. Statistically defensible, psychologically difficult: most users switch the robot off during a flat period, usually just before it picks up again.
Your time horizon. A six-month track record tells you almost nothing. On a trending asset, a mediocre strategy can post six excellent months.
A generic ranking ignores all four parameters. That is why it is of no use to you.
Step 1 — Demand a real track record, not a backtest
A backtest is a simulation run on past data. It can be tweaked until it produces the desired curve, and doing so is the most routine operation on the EA market.
What counts is live trading: a real account, with slippage, variable spreads and order rejections. Failing that, a demo account tracked over a meaningful period is still far better than a backtest.
Three questions to put to the vendor:
- Since what date has the robot been running in live conditions?
- At which broker, on what account type, with what average spread?
- Can I have the full statement, rather than a screenshot?
A refusal on the third question ends the discussion.
Step 2 — Read the drawdown before the return
The advertised return is a sales argument. Maximum drawdown is survival information: it is the worst equity decline from a peak, and therefore what you will have to endure without intervening.
| Maximum drawdown | Interpretation |
|---|---|
| < 10% | Comfortable, workable on most account sizes |
| 10 – 20% | Acceptable if the track record is long and leverage is controlled |
| 20 – 35% | Reserved for dedicated capital you accept seeing tied up |
| > 35% | Ruin becomes a likely scenario, not a hypothesis |
One essential clarification: drawdown calculated on closed positions is a floor. A robot that lets losing positions run shows a flattering closed drawdown and a far worse floating drawdown. That is precisely the profile of grid strategies.
Step 3 — Check that win rate and payoff are consistent
A 90% win rate is impressive and proves nothing. It has to be read alongside the payoff (average win ÷ average loss).
- High win rate and payoff above 1: a rare and solid profile.
- High win rate and payoff below 0.3: performance rests entirely on the frequency of wins. One unusual losing streak wipes out several months.
- Low win rate and high payoff: a trend-following profile, psychologically demanding but structurally sound.
The second case describes the majority of robots sold with perfectly smooth curves.
Step 4 — Look explicitly for martingale behavior
A robot that increases position size after a loss is running a martingale. The mechanism produces a steady curve for months, then a total loss on a single unfavourable sequence.
Signs you can spot in a statement:
- lot sizes that double after every losing trade;
- several simultaneous positions on the same asset, in the same direction;
- regular small gains, interrupted by a loss an order of magnitude larger;
- the absence of a stop loss.
A vendor will call this "adaptive position management" or "recovery". The vocabulary changes; the risk of ruin does not.
Step 5 — Test the robot under your own conditions
An EA that has passed the previous four steps still has to be tested at your broker, on a demo account funded with the amount you actually intend to commit, for at least a month. That is when you verify three things the vendor's statement will never tell you: the effective spread, the slippage on execution, and your own reaction to a losing streak.
The signals that should end the evaluation
- A guaranteed monthly return, whatever the figure.
- No drawdown disclosed, or a drawdown below 3% alongside a high return.
- A track record shorter than six months presented as proof.
- Refusal to provide the raw statement.
- Sales pressure: limited offer, number of licenses remaining, countdown timer.
Evaluating a robot in practice
This method comes down to reading a statement rigorously. That is exactly what Judgebot automates: you import your MT4/MT5 statement, and the tool computes the real drawdown, the profit factor, the recovery factor and the payoff, detects martingale and grid behavior, then assigns an A/B/C/D grade along with the weaknesses it identified.
Nothing is taken on trust: everything is recalculated from your trades.
👉 Analyze your statement for free
You can also browse the Expert Advisor directory, which lists the robots available on the market with their actual verification level, and states explicitly when no auditable track record exists.
In summary
The best robot is not the one showing the highest return, but the one whose risk is compatible with your capital, your broker and your patience. Demand a real track record, read the drawdown before the return, cross-check win rate against payoff, rule out any martingale, then test it under your own conditions. A robot that survives these five steps is worth considering. The others do not deserve your money, whatever their ranking.