How to evaluate an expert advisor before you trade it

Most expert advisors look good in the screenshot that sells them. Evaluating one means looking past the headline return to the evidence underneath: how many trades, over what period, with what drawdown and what money management. This guide covers the numbers worth checking, the red flags that should end your interest and the tests you can run yourself.

Know what evidence you're looking at

EA results come in three forms, and they aren't equally convincing.

The best case combines all three and shows they agree. Compare like with like: the same settings, symbols and risk level. Live results usually trail a backtest a little because of costs and execution, and that is normal. A large or growing gap is not. If a backtest shows a 10% maximum drawdown and the live account has already been through 30%, the backtest doesn't describe the EA you would be running.

Read a live account's history, not just its curve. Note the start date, the deposits and withdrawals, and whether the account was ever restarted. Large deposits can hide losses in a growth chart, and a track record that begins again after every bad patch only shows the survivors.

Is the sample big enough?

A handful of trades can look like anything. A short run of winners proves little, especially for systems built to win often in small amounts. As a rough guide, look for:

• Hundreds of trades, not dozens. The more trades behind a result, the less it depends on luck.
• Several years of data. The EA should have met different market regimes: strong trends, long ranges, quiet stretches and sharp volatility.
• Results by year or month. A breakdown shows whether one exceptional stretch is carrying the total.
• A longer period for slow EAs. An EA that trades twice a month needs many years to build a meaningful sample.

A swing EA with 60 trades over five years can still be sound, but its numbers carry a wide margin of error. Size it more cautiously than an EA with thousands of trades behind it, and expect its live results to wander further from the test.

Drawdown: look at equity, not just balance

Drawdown is the fall from a peak to a low point. Balance drawdown counts only closed trades. Equity drawdown includes the floating loss on trades that are still open. An EA that never closes a losing trade can show almost no balance drawdown while its equity falls by 40%. Always ask for the maximum equity drawdown, and in the strategy tester read the equity line, not just the balance line.

Drawdown also costs more to recover than it takes to lose. A 20% loss needs a 25% gain to get back to where you started, and a 50% loss needs a 100% gain. The drawdown calculator shows this for any figure.

Plan for live drawdown to exceed the historical maximum at some point. The worst drawdown in a test is the worst one so far, not the worst one possible.

Profit factor and the other key numbers

No single figure proves an EA is robust, but together they show how it makes its money. The ranges below are rough guides for large samples, not hard rules.

The win rate deserves its own warning. A system that wins 95% of the time sounds safe, but if each win is 5 pips and each loss is 150 pips, it loses money. Over 100 trades it wins 475 pips and loses 750. Very high win rates usually come from wide or missing stop losses, averaging down or grids.

To see how win rate, reward to risk and risk per trade combine, try the risk of ruin calculator. It estimates how likely a strategy with those figures is to hit a drawdown limit you set.

Account for spread, slippage and commission

Trading costs hit EAs with small average profits hardest. Work out what the EA makes per trade before costs and compare it with what each trade costs.

Say an EA averages 4 pips per trade before costs. On an account with a 1-pip spread and commission worth 0.7 pips, each trade costs about 1.7 pips, and slippage on entry and exit might add a few tenths more. Close to half the edge goes to costs. On an account with a 2-pip spread, very little is left. An EA that averages 40 pips per trade barely notices the same costs.

Ask whether published results include commission and a realistic spread. Backtests run with a spread of 0 or 1 point, or with commission left out, flatter scalpers badly. Check when the EA trades, too. EAs that trade at rollover or around news releases meet the widest spreads of the day.

Red flags that should end your interest

• No stop loss. Some EAs close losers only by their own logic, or set a stop so wide it never triggers. A single bad move then has no limit.
• A hidden grid or martingale. Look in the trade history for several trades in the same direction opened at regular price intervals, lot sizes that grow after losses, and groups of trades closed at the same moment. Our grid and martingale guide explains the risk.
• Averaging down. Adding to a losing position lowers its break-even price but raises exposure exactly when the trade is going wrong.
• A near-perfect win rate. Tiny wins, rare huge losses and a win rate close to 100% describe a fragile system.
• Promised returns. Phrases like “10% a month” or “never loses” are marketing, not trading. Nobody can promise returns.
• A short history. A few weeks or months of live trading, especially in a single kind of market.
• Account resets. Several accounts started at different times with only the best one shown, or a track record that restarts after each blow-up.
• Backtests only. A perfectly straight equity curve with no live account to compare it with.
• Unclear lot sizing. Risk settings that aren't documented, or lots that grow with the balance in a way you can't control.

Stress test it yourself

If you can run the EA in the strategy tester, a few extra tests tell you more than the developer's report. Set up the tester as described in our backtesting guide, then:

1. Raise the spread. Repeat the test with a spread 50% to 100% wider than your account's typical spread. A robust EA loses some profit. A fragile one turns negative.
2. Change the start date. Start a few months earlier or later. A result that depends on one start date often depends on one lucky trade.
3. Test other periods. Run years the developer didn't show, especially strong trends and long, quiet ranges.
4. Try related symbols. A general trend EA built on EURUSD should behave sensibly on similar pairs, even if it isn't profitable on all of them. EAs built for a single symbol are the exception.
5. Nudge the inputs. Move key settings a small step either way. Profit that vanishes after a small change suggests overfitting.
6. Add delays. In MT5, set Delays to a random delay to see how sensitive the EA is to slower execution.
7. Forward test on demo. Run it for a few weeks with your own broker before any live money is involved.

Keep a simple log of each run: settings, period, spread, net profit, profit factor and maximum equity drawdown. Patterns show up quickly once the results sit side by side.

Check the developer, support and updates

When you run an EA, you also rely on whoever maintains it. Before you commit, check for:

• Documentation that explains the strategy in plain terms: what triggers a trade, how it exits, and whether it uses grids, martingale or hedging.
• Recommended symbols, timeframes and minimum balance, ideally with preset .set files.
• An update history with dated changes, including fixes for new MetaTrader builds.
• Support you can actually reach, with answers that address your question rather than repeating sales copy.
• Clear license terms: how many accounts, live or demo, and what happens when you change broker or VPS.

Use these checks when you browse our catalog. If nothing fits your rules, our custom development service builds an EA to your specification, so you know exactly what it does and why.