A profitable backtest does not automatically mean an Expert Advisor is safe.
Many EAs can produce attractive growth curves for months or even years while quietly accumulating risk. The danger may only become visible during an unusually strong trend, a volatility spike, widening spreads, or a long sequence of losing trades.
No EA can be guaranteed never to lose an account. However, buyers can identify many warning signs before risking real money.
This article explains what to check before purchasing, renting, or running an EA.
1. Understand How the EA Makes Money
Before looking at profit, ask one basic question:
What happens when the first trade is wrong?
A transparent strategy should have a clear answer.
For example, does the EA:
- Close the trade at a defined stop loss?
- Open another trade in the same direction?
- Increase the next lot size?
- Hold the position until the market returns?
- Add positions at fixed price intervals?
- Hedge the original position?
- Close all trades at a maximum equity loss?
You do not need the full source code, but you should understand the general trading and recovery logic.
An EA becomes dangerous when its entire survival depends on the assumption that the market must eventually reverse.
2. Check for Martingale Lot Progression
Martingale usually means increasing position size after a losing trade or as the market moves against existing positions.
A sequence might look like this:
- 0.01 lot
- 0.02 lot
- 0.04 lot
- 0.08 lot
- 0.16 lot
- 0.32 lot
The early trades appear small, but total exposure grows rapidly.
After only six entries, the combined position size is already:
0.63 lots
The problem is not simply that the EA opens multiple trades. The problem is that exposure may grow faster than the account can tolerate.
Before using the EA, check:
- Does the lot size increase after losses?
- What is the lot multiplier?
- Is there a maximum lot size?
- Is there a maximum number of recovery trades?
- Can the progression continue indefinitely?
- What happens when the maximum level is reached?
A martingale EA can recover many losing sequences before one extended market move causes a catastrophic loss.
A high historical win rate does not remove this risk.
3. Check Whether the EA Uses Grid Trading
A grid EA opens multiple positions as price moves through predefined levels.
Grid trading is not automatically unsafe. The risk depends on how exposure is controlled.
A more controlled grid may include:
- A fixed maximum number of positions
- Small and consistent lot sizes
- A basket stop loss
- A maximum account drawdown limit
- A maximum total exposure
- A volatility filter
- A defined trading direction
A dangerous grid may continue opening trades without a meaningful limit.
When reviewing a grid EA, ask:
- How far apart are the grid entries?
- Does the spacing change with volatility?
- Are lot sizes fixed or increasing?
- How many positions can be opened?
- Is there a maximum basket loss?
- Can the EA continue adding trades during a major trend?
A grid that appears stable in a ranging market may behave very differently during a prolonged one-directional move.
4. Prioritize Equity Drawdown Over Balance Drawdown
Balance reflects closed trades.
Equity includes both closed trades and the floating profit or loss of open positions.
This distinction is extremely important.
An EA may show a smooth rising balance while holding large floating losses. The balance curve looks healthy because the losing trades have not yet been closed.
Suppose an account shows:
- Balance: $10,000
- Floating loss: $3,000
- Equity: $7,000
The account is already down 30% in real terms, even though the balance has not changed.
When reviewing a backtest or live signal:
- Compare the balance and equity curves
- Look for large gaps between them
- Check the maximum equity drawdown
- Check how long floating losses remain open
- Look for unresolved positions near the end of the test
A smooth balance curve with a deeply fluctuating equity curve is a major warning sign.
5. Confirm That the EA Uses a Real Loss Limit
A stop-loss input does not necessarily mean every trade is protected.
Some EAs use:
- Individual trade stop losses
- Hidden stop losses
- Basket stop losses
- Equity-based stop losses
- Time-based exits
- No fixed stop loss at all
Ask what actually closes a losing position.
A protective system should define a point where the EA accepts that the trade or trading sequence is wrong.
Important questions include:
- Is a stop loss placed on every trade?
- Is the stop loss visible at the broker?
- Can the EA remove or widen the stop?
- Is there a basket-level stop?
- Is there a daily loss limit?
- Is there a maximum equity drawdown setting?
- Does the protection still work if the terminal or VPS disconnects?
A server-side stop loss may remain active during a terminal outage. A protection feature controlled only by the EA may require the platform to remain connected.
6. Look at the Worst Trade, Not Only the Average Trade
Average profit and win rate can be misleading.
Consider an EA that produces:
- 100 winning trades of $10 each
- 1 losing trade of $1,200
It has a win rate above 99%, but the strategy is still unprofitable.
This pattern is common in systems that take small profits while allowing losses to grow.
Check:
- Largest winning trade
- Largest losing trade
- Average winning trade
- Average losing trade
- Profit factor
- Consecutive losses
- Maximum floating loss
- Maximum trade duration
Pay particular attention when the largest loss is many times larger than the average win.
A system should not depend on maintaining an almost perfect win rate to survive.
7. Review Maximum Exposure
The risk of one trade is not the same as the risk of the entire account.
An EA may risk only 1% on the first trade but later open ten additional positions.
You should evaluate total exposure across all open trades.
Check:
- Maximum number of simultaneous positions
- Maximum total lot size
- Number of symbols traded at once
- Whether positions are correlated
- Maximum margin usage
- Maximum risk per trading sequence
- Maximum risk across the entire portfolio
For example, three EAs trading gold in the same direction may behave like one highly leveraged strategy rather than three independent systems.
Portfolio risk should be measured at account level, not EA level alone.
8. Check the Margin Level During the Worst Period
A strategy may survive its historical test while coming dangerously close to a margin call.
Look beyond net profit and drawdown.
Check:
- Minimum margin level
- Maximum margin used
- Free margin during the worst floating loss
- Leverage used in the test
- Whether the account could survive wider spreads
- Whether the broker’s stop-out level was considered
An EA that reaches extremely low margin levels has little room for:
- Slippage
- Spread expansion
- Price gaps
- Swap charges
- Execution delays
- Different broker specifications
A small difference between the backtest and live environment may be enough to cause forced liquidation.
9. Test a Larger Starting Capital and a Smaller Starting Capital
An EA may perform well only because the test used a large deposit relative to its lot size.
Test the system with different account balances.
For example:
- $1,000
- $3,000
- $5,000
- $10,000
Keep the risk logic realistic.
This helps reveal whether the strategy:
- Scales correctly
- Uses a hidden minimum lot
- Becomes too aggressive on small accounts
- Depends on large free margin
- Calculates risk based on balance correctly
- Opens excessive volume after a drawdown
A strategy that survives on $10,000 may fail quickly on $1,000 if the minimum tradable volume prevents proper risk scaling.
10. Reduce the Risk and Compare the Results
A robust EA should normally remain profitable when risk is reduced.
Try lowering:
- Risk per trade
- Fixed lot size
- Maximum simultaneous trades
- Lot multiplier
- Recovery levels
- Maximum portfolio exposure
If the strategy stops working when risk is reduced slightly, the original results may depend heavily on aggressive position sizing.
The goal is not to maximize the backtest profit.
The goal is to find a risk level that the account can realistically survive.
11. Run the Test Across Different Market Conditions
Do not judge an EA from a short test covering only one favourable period.
Include periods with:
- Strong trends
- Low volatility
- High volatility
- Market crashes
- Sudden reversals
- Wide spreads
- Long ranging conditions
- Major economic events
- Changing interest-rate environments
An EA tested only during its ideal market condition may fail when the environment changes.
A longer test does not guarantee safety, but it gives the strategy more opportunities to encounter difficult conditions.
12. Check Whether the Backtest Ends With Open Trades
Some backtests look profitable because losing positions remain open when the test ends.
The report may show a high balance while the account still has a large floating loss.
Always inspect:
- Open trades at the end of the test
- Final account equity
- Unrealized profit or loss
- Number of unresolved recovery positions
- Whether the test was stopped during a favourable moment
A fair test should not hide risk by ending before the trading cycle is completed.
13. Use Realistic Testing Conditions
An EA may look safer than it is when tested with ideal conditions.
Include realistic:
- Spreads
- Commission
- Swap
- Slippage assumptions
- Tick data
- Symbol specifications
- Trading hours
- Broker execution conditions
This is particularly important for:
- Scalping EAs
- News-trading EAs
- High-frequency systems
- Strategies with tight stop losses
- Systems trading during rollover
- EAs that open many positions
A few extra points of spread may transform a profitable strategy into a losing one.
14. Compare Backtests With Live Results
Backtests show how a strategy might have behaved under simulated conditions.
Live results show how it performs with real execution.
When reviewing a live signal, check:
- Account age
- Number of trades
- Maximum equity drawdown
- Average monthly growth
- Deposit load
- Trading frequency
- Open positions
- Floating profit or loss
- Lot-size progression
- Balance and equity curves
Be cautious when the live account is very new.
A recovery-based strategy may look excellent before it encounters its first major losing cycle.
Several profitable months are encouraging, but they do not prove that the strategy can survive every market condition.
15. Examine the Relationship Between Return and Drawdown
High returns are usually associated with higher risk.
Be cautious when an EA claims to produce extremely high returns with almost no drawdown.
Compare the return with:
- Maximum drawdown
- Recovery factor
- Margin usage
- Total exposure
- Test duration
- Number of trades
For example:
EA A
- Return: 40%
- Maximum equity drawdown: 10%
EA B
- Return: 80%
- Maximum equity drawdown: 65%
EA B produced more profit, but the account experienced much greater risk.
A 65% drawdown requires a gain of approximately 186% to return to the previous account peak.
The deepest drawdowns are much harder to recover from than they first appear.
16. Perform a Simple Stress Test
A backtest should not be treated as a prediction.
Try making the conditions worse.
For example:
- Increase the spread
- Add commission
- Use a later starting date
- Change the broker
- Change the timeframe
- Reduce the initial deposit
- Increase slippage assumptions
- Start the test before a major market shock
- Remove the most profitable year
- Test a nearby parameter value
If a small change causes the EA to collapse, the strategy may be overoptimized or highly sensitive.
A more robust system should not depend on one exact parameter combination.
17. Be Careful With Extremely High Win Rates
A 95% or 99% win rate may look impressive, but it can also indicate that losses are being delayed rather than controlled.
Some EAs generate:
- Frequent small wins
- Long-held losing positions
- Rare but very large losses
Ask:
- How large is the average loss compared with the average win?
- How long are losing trades held?
- Are losing positions averaged?
- Does the EA close losses, or wait indefinitely?
- Has the strategy experienced a complete recovery cycle?
Win rate should never be evaluated without loss size and drawdown.
18. Check the EA’s Emergency Controls
A safer EA should ideally include clear account-protection tools.
Useful controls may include:
- Maximum daily loss
- Maximum account drawdown
- Maximum floating loss
- Maximum number of trades
- Maximum lot size
- Maximum total exposure
- Spread filter
- Slippage protection
- Trading-session filter
- News filter
- Friday closing option
- Automatic recovery shutdown
- Minimum margin-level protection
These features do not guarantee safety, but they can limit how much damage one abnormal event causes.
Most importantly, verify that the protections are enabled in the settings you actually plan to use.
19. Calculate the Risk in Money
A percentage can feel abstract.
Convert the expected drawdown into account currency.
For example, on a $10,000 account:
- 10% drawdown = $1,000
- 20% drawdown = $2,000
- 30% drawdown = $3,000
- 50% drawdown = $5,000
Then ask:
Would I continue running the EA after losing this amount?
Many traders choose aggressive settings during a backtest but stop the EA emotionally after the first major live drawdown.
Use a risk level you can realistically tolerate, not merely one that produces the highest historical profit.
20. Never Assume Historical Drawdown Is the Maximum Possible Drawdown
If a backtest shows a maximum drawdown of 15%, this does not mean the EA can never exceed 15%.
Future drawdown may be higher because:
- Market conditions change
- The historical sample is limited
- Live execution is worse
- Volatility increases
- Correlations change
- A rare losing sequence occurs
- The strategy was overoptimized
It is sensible to maintain a safety margin.
An account should not be funded so tightly that a drawdown slightly above the historical maximum causes a margin call.
Warning Signs That Require Extra Caution
Be especially careful when an EA has several of the following characteristics:
- No visible stop loss
- Increasing lot sizes
- Unlimited grid levels
- Extremely high win rate
- Very small average wins
- Rare but enormous losses
- Large gap between balance and equity
- Long-held losing trades
- High deposit load
- Low minimum margin level
- Short backtest period
- Very small trade sample
- Backtest ending with open positions
- Perfect-looking balance curve
- Extremely high advertised returns
- No verified live performance
- Results that collapse when settings are changed slightly
One warning sign alone does not automatically make an EA bad.
Several warning signs together should not be ignored.
Practical EA Safety Checklist
Before running an EA with real money, confirm that you know:
✅ How the EA handles a losing trade
✅ Whether it uses martingale or grid logic
✅ The maximum number of open positions
✅ The maximum possible lot size
✅ The maximum total exposure
✅ The maximum historical equity drawdown
✅ The longest drawdown duration
✅ The lowest margin level
✅ Whether every trade or basket has a loss limit
✅ Whether emergency equity protection is available
✅ Whether the backtest used realistic costs
✅ Whether the test covers difficult market conditions
✅ Whether the live results support the backtest
✅ Whether the account can survive a larger-than-expected drawdown
✅ Whether the risk settings match your account size
Final Thoughts
There is no test that can prove an EA will never blow an account.
Markets can produce conditions that have never appeared in the historical data. Broker execution can change, spreads can widen, and strategies can stop performing as expected.
However, most catastrophic EA failures are not completely unpredictable.
The warning signs are often visible:
- Excessive leverage
- Unlimited recovery trades
- Increasing lot sizes
- Large floating losses
- Weak margin protection
- No defined exit point
- Aggressive risk settings
Do not purchase an EA based only on net profit, win rate, or a smooth balance curve.
Study the equity curve. Understand the worst-case exposure. Test difficult periods. Reduce the risk. Confirm how losses are controlled.
The most important question is not:
“How much can this EA make?”
It is:
“What can happen to my account when the strategy is wrong?”
Protecting capital begins with understanding the risk before the first live trade is opened.
