How to test a free EA on a demo account
Free EAs are unverified by definition — so the rule is simple: demo first, for at least a few weeks, with realistic lot sizes. Here is the routine that separates a usable robot from a broken or dangerous one.
1. Open a demo, not a cent-account
File → Open an Account → pick your broker's demo server → Demo account. Choose leverage and balance similar to what you'd really trade with; a $100,000 demo tells you nothing about a $500 live account. Use the same account type (hedging vs netting on MT5) you'd use live — some EAs refuse to run on netting accounts.
2. Attach it correctly
Install as in the install guide. Put it on the symbol and timeframe the EA is meant for (check the file name and the Telegram post). If it uses several pairs it will usually say so in the inputs.
3. The inputs that matter
- Lot / Risk % — start tiny (0.01 lot or 0.5–1% risk). A robot that only "works" at big lot sizes is a martingale in disguise.
- Magic number — a unique ID so the EA only manages its own trades. Use different numbers if you run several EAs.
- Max spread / slippage — scalpers die on wide spreads; set a realistic limit for your broker.
- Trading hours / news filter — many gold and scalping EAs behave completely differently at the open or during news.
- Take profit / stop loss — if there is no stop-loss input at all, be extra careful: grid and martingale systems often trade without one.
4. Read the Experts and Journal tabs
At the bottom of the terminal (Toolbox in MT5): the Experts tab shows what the EA prints — initialisation messages, errors like "invalid volume", "not enough money", "hedging prohibited"; the Journal tab shows terminal-level events. Two things to look for: (a) does it initialise without errors, (b) does it actually open and close trades over a few days? Silence for a week usually means it's waiting for a condition, a wrong symbol, or a licence check that failed.
5. Strategy Tester in ten minutes
- Open the tester (View → Strategy Tester, or Ctrl+R).
- Choose the EA, symbol, timeframe and a date range of at least 6–12 months. Modelling: Every tick (MT4) / Every tick based on real ticks (MT5) for anything that scalps.
- Set the same inputs you'd use on demo, then Start.
- Read the report: Profit factor (above ~1.3 is decent), Maximal drawdown (how deep the equity dived), Total trades (fewer than ~100 proves nothing), and the equity curve — smooth and rising beats a hockey-stick from one lucky month.
Backtests are not proof — spreads, slippage and broker feeds differ — but a robot that loses in the tester will not magically win live.
6. What "passing" looks like
Three to four weeks on demo with: no unexplained errors, drawdown you could stomach on live money, trade frequency that matches the description, and results in the same ballpark as the backtest. Then go live small, ideally on a VPS so the EA runs 24/5, and keep the demo running alongside for comparison.
Browse robots to test: MT5 EAs · MT4 EAs · gold EAs. Trouble getting one to trade? EA not working — the usual fixes.
Frequently asked questions
How long should I test an EA on demo?
At least 2–4 weeks of forward testing, plus a 6–12-month Strategy Tester run. Short tests only capture one market mood.
Do backtests prove an EA is profitable?
No. They rule out broken robots and show the drawdown profile, but live spreads, slippage and broker feeds differ. Treat a good backtest as a permission slip for demo, not for live money.
What lot size should I use when testing?
The smallest available (usually 0.01) or 0.5–1% risk per trade. If the EA only makes sense at large lots, it is relying on martingale-style recovery.