Are free forex EAs safe to use?

Free does not mean safe or unsafe. A free open-source EA has one big advantage: you can read its code and see how it trades. Its risks are the same as any EA's, such as grid or martingale sizing, missing stop-losses and over-fitted settings. Download only source files, never installers, and test on a demo account before risking real money.

Is a paid EA better than a free one?

Not automatically. A price tag says nothing about the strategy, and many paid robots come only as compiled files, so you cannot check what they do. Judge a paid EA by the same evidence as a free one: long backtests, a demo run and verified live history that shows drawdown. If the seller will not show that, the price buys you nothing.

How much money do I need to run a forex EA?

It depends on the EA's lot size and its worst drawdown, not on a round number. Take the largest drawdown from a realistic backtest, decide what share of your account you could lose without panicking, and divide. For example, a $500 worst case that you want to keep under a quarter of the account needs at least $2,000.

Can a forex EA be profitable in the long run?

Some EAs make money for a period, but none does so in every market. An EA's rules suit certain conditions, and when the market changes character its results can turn negative for months or never recover. That is why you need a switch-off rule, a cap on risk and regular reviews against the EA's tested behaviour.

How to Choose a Safe Forex EA: A 12-Point Checklist

Intermediate8 min read

Magnifying glass over the gears of a small mechanism — Image by Shane Aldendorff on Unsplash

How do you choose a safe forex EA?

Choose an EA by checking four things: its source (readable code, a licence and an active author), its logic (no hidden grid or martingale, a stop-loss, a cap on trades), its evidence (long tests, a demo run and real history) and its fit with your account and broker. Skip any EA that fails a check you cannot explain.

Diagram of a 12-point checklist for choosing an expert advisor, grouped into the source, the logic, the evidence and the fit — The whole checklist on one page. Work left to right: if an EA fails a source or logic check, there is no need to spend weeks on the evidence.

An EA, or expert advisor, is a program that trades for you inside MetaTrader. The 12 points below work for free open-source EAs and paid ones alike. Score each as you go; the scoring method is at the end.

Is there a best forex EA?

No. There is no universal best EA, and anyone selling one is selling a story. Every EA is a fixed set of rules, and every set of rules suits some markets and fails in others. A trend-following EA does well in long moves and bleeds in choppy ranges; a range EA does the opposite. Results also depend on your broker's costs, your account size and how much drawdown, the fall from a peak in your account, you can sit through. The useful question is not “which EA is best?” but “which EA fits my account, and how does it fail?”

The source: can you see what it does? (points 1-3)

1. Readable code: an open-source EA ships its .mq4 or .mq5 source file, the text you can read and compile. A compiled .ex4 or .ex5 file alone is a sealed box. Even without coding skills you can search the source for its settings and for how lot sizes change, as our guide to open-source EAs on GitHub shows.
2. A licence: the licence says what you may do with the code. MIT, BSD, Apache and GPL are open-source licences; no licence means the author keeps all rights. You can list only EAs with an open-source licence.
3. Signs of life: check the last update, whether the author has archived the project (frozen it as read-only), what else the author has published, and the Issues tab on GitHub, where users report bugs. Code untouched for years may not handle newer platform builds or broker changes. The library can show EAs updated in the last six months.

The logic: how does it trade, and how does it lose? (points 4-6)

4. Type and strategy: is it a trading robot that opens trades, or a trade manager that handles yours? Is it trend following, scalping, breakout or mean reversion? Each has a normal way of losing, explained in our guide to types of expert advisors.
5. No hidden grid or martingale: a grid opens a ladder of trades at fixed distances; a martingale raises the lot size after losses. Look for a lot multiplier above 1, a grid step, or “averaging”, adding to a losing trade to improve its average price. With a multiplier of 2, the sixth trade is 2 × 2 × 2 × 2 × 2 = 32 times the first. Our guide to martingale and grid trading shows why these look smooth until one strong trend.
6. A stop-loss and a cap on trades: a stop-loss is an order that closes a losing trade at a set price. Check that the EA sets one on every trade and that an input limits the number of open trades. With neither, its loss has no ceiling.

The evidence: what has it actually done? (points 7-9)

7. A long backtest at a realistic spread: many years, the most accurate modelling mode, your real deposit and a spread at least as wide as your broker's. Run it yourself rather than trusting a screenshot; our guide on how to backtest an EA covers the settings.
8. A forward demo test: weeks or months on a demo account, which uses virtual money, under live prices. Results far worse than the backtest point to costs, slippage or over-fitting.
9. Verified live history, and plenty of trades: the strongest evidence is a real-money account tracked by an independent service, showing equity drawdown, not just balance, over hundreds of trades. A handful proves little: for an EA that wins 9 trades in 10, any run of ten trades is all winners about a third of the time (0.9 multiplied by itself ten times ≈ 0.35).

The fit: does it suit your account and broker? (points 10-12)

10. Platform, symbol and timeframe: an MT4 EA will not run on MT5, or the other way round. Use the symbol and timeframe it was built and tested on; an EA tuned for EURUSD on H1 may behave very differently on gold or M5.
11. Minimum deposit versus your account: the drawdown needs room. Suppose the backtest's worst drawdown was $600 at 0.10 lots and you want that worst case under 20% of your account: you need 600 ÷ 0.20 = $3,000. On a $1,000 account, cutting the lot to 0.03 scales the drawdown to about $180, or 18%.
12. Broker conditions: check the spread, the commission and the stops level, the minimum distance from price at which the broker accepts a stop-loss or take-profit. Right-click the symbol in Market Watch and choose Specification to see it. A scalper that needs a 30-point stop fails where the stops level is 50 points. Account types differ here; see types of forex brokers, and compare brokers if your costs are high.

Forex EA red flags: is this EA a scam?

Commercial robots and signal sellers reuse the same tricks. Walk away if you see:

Guaranteed returns: “10% a month” is a sales line, not a forecast.
“No losses”: a near-perfect win rate usually means losers are held open, not avoided.
A “recovery” mode: often a polite name for martingale. Ask how the lot size changes after a loss.
Balance-only proof: screenshots with no equity drawdown and no open trades.
Broker pressure: you must deposit with one specific broker, often unregulated, or hand over your login.
Only compiled files, or a download that asks you to turn off your antivirus.

Our guide to forex scams explains how these schemes work and how to report them.

How the EA Library's badges and filters help

The ForexR EA Library lists open-source MT4 and MT5 EAs from GitHub and links each to its original repository. It checks repositories automatically and leaves out ones that ship executables, ask you to disable antivirus, hide files behind archive passwords or show signs of bought stars. Its summaries and tags are written automatically from each README and its code. It does not test, verify or recommend any EA, so the checklist is still your job. It does make several points quicker:

High risk badge: marks grid or martingale EAs (point 5), and the “Hide grid and martingale” toggle lets you browse without them.
Archived by its author badge: the project is frozen (point 3).
Licence field: shown with stars, forks and last update on each EA page (points 2 and 3).
Input parameters table: the EA's settings and defaults, read from its code, where multipliers and stop-loss inputs show up (points 5 and 6).
“Updated in 6 months” toggle: linked above, for active projects.

Screenshot of a grid-trading EA page in the EA Library showing a High risk badge, a last update eight years ago and a GPL-3.0 licence — Two warnings at a glance on a real library page (screenshot, September 2026): a High risk badge for grid logic and a last update eight years ago. The licence tile shows it is open source.

A simple EA scoring method

Score each of the 12 points 2 if it clearly passes, 1 if you are unsure and 0 if it fails. The maximum is 12 × 2 = 24.

20-24: worth your own backtest and a demo run.
14-19: close the gaps first, for example by reading the code.
0-13: skip it, or keep it only as something to learn from.
Deal-breakers: an uncapped martingale or grid, or no stop-loss at all, is a fail whatever the total.

Example: an EA scores 2 on nine points, 1 on two and 0 on one, so 18 + 2 + 0 = 20. If that 0 is the stop-loss check, it is still a no.

A checklist screens out the worst EAs, but it cannot make any EA safe or tell you how it will trade next month. Automated trading of forex and CFDs carries a high risk of loss, so only risk money you can afford to lose.