Evaluating EA Drawdown Risks

Drawdown is the single most important number to understand before entrusting any capital to an Expert Advisor. A high headline return means little if the equity curve dropped 60% along the way. This guide defines drawdown properly, explains the different types you will see on backtest and live reports, and shows how to judge whether an EA is genuinely survivable or simply lucky so far, so that you can compare EAs on the metric that actually determines whether you can hold them through a rough period.

What Drawdown Actually Means

Drawdown is the percentage decline from a peak in account equity to a subsequent trough, before a new equity high is made. If an account grows from £10,000 to £12,000 and then falls to £9,600 before recovering, the maximum drawdown for that period is 20% (£2,400 / £12,000).

Drawdown is different from a loss. A loss is on a single trade. Drawdown is the peak-to-trough decline of the whole account, which is what matters if you plan to hold an EA through more than a few weeks of live trading.

The Types of Drawdown You Will See

Different reports use different definitions. Confusing them is one of the most common EA evaluation mistakes.

Term	Definition
Maximum drawdown	Largest peak-to-trough decline observed in the report period
Average drawdown	Mean of all recorded drawdown periods
Relative drawdown	Maximum drawdown expressed as a percentage of the equity peak
Absolute drawdown	Largest decline from initial deposit, not from later peaks
Balance drawdown	Based on closed trades only
Equity drawdown	Based on floating (open) positions included

Equity drawdown is usually more meaningful than balance drawdown for grid, martingale, or hedging EAs, because those strategies often carry deep floating losses that never show up in the balance figure until a basket is finally closed.

Why Backtest Drawdown Is a Floor, Not a Ceiling

The drawdown you see in a MetaTrader Strategy Tester report is the worst period the EA experienced against the historical data it was optimised on. In live trading, the worst period is almost always deeper, because:

• Spreads and slippage are worse than modelled.
• Volatility regimes not present in the sample can appear.
• Execution delays create fills the tester never simulated.
• Optimisation bias flatters the sample.
• Broker requotes and rejections do not appear in the tester.

As a rough rule, live maximum drawdown often reaches 1.5x to 2x the backtest figure over a long enough period. This is why the MetaTrader 5 Strategy Tester documentation stresses the importance of proper history data and realistic conditions.

How to Evaluate an EA’s Drawdown Honestly

1. Compare drawdown to expected annual return. A ratio of return-to-max-drawdown below 1 is usually poor. Above 2 is respectable.
2. Look at the depth AND the duration. A 20% drawdown that lasts three days is very different from a 20% drawdown that lasts nine months.
3. Check the recovery pattern. Grid EAs often “recover” only because they refuse to close losing baskets. That is not a recovery, that is delayed acknowledgement.
4. Ask what leverage produced the figure. A 15% drawdown at 1:500 leverage is not the same as 15% at 1:30.
5. Cross-reference with the equity curve, not just the number. A single deep drawdown late in the period is more alarming than a stable series of small ones.
6. Look at drawdown frequency. An EA that drops 10% twice a year is very different from one that drops 10% every month.

Position Sizing and Perceived Drawdown

Drawdown scales with risk per trade. Doubling lot sizes roughly doubles drawdown. This is why claims like “only 5% drawdown” are meaningless without knowing the position sizing used to generate them. Many EAs allow the user to tune lot size, and users often push it beyond what the strategy can sustain. If you see published examples using 0.01 lots on a £10,000 account, that is roughly 0.1% risk per trade, not the risk a typical retail user will actually apply.

When comparing published drawdown between two EAs, normalise for risk per trade. A 12% figure at 1% risk per trade is a very different signal from a 12% figure at 0.1% risk per trade.

Structural Drawdown Traps

Some EA architectures produce a specific drawdown shape that misleads casual evaluators.

• Martingale doubling produces a long calm curve punctuated by rare catastrophic drops.
• Grid without a stop shows an ever-growing floating loss that is not counted in balance drawdown.
• Scalpers on wrong brokers show huge drawdown because spreads eat most trades.
• News EAs can show near-zero drawdown for months and then a single catastrophic event.
• Averaging-in strategies hide drawdown by adding to losers until the market reverses.

Our editorial view on vetted EA due diligence explains the checks we apply before considering any EA fit for publication.

Setting a Personal Drawdown Threshold

Before deploying any EA, set two numbers:

• The drawdown at which you will manually intervene or pause the EA.
• The drawdown at which you will consider the EA broken and stop it entirely.

Write them down. Automation is only useful if you also automate discipline. A written threshold decided in a calm week is almost always a better guide than an in-the-moment decision made during a losing period.

The Relationship Between Drawdown and Time in Market

A short backtest with low drawdown is almost meaningless. Any strategy will show a low drawdown over a lucky sample. What matters is how the EA behaved through:

• Multiple central bank cycles
• Both high and low volatility regimes
• Major geopolitical or policy shocks
• Extended range-bound periods

An EA that only performed during a specific low-volatility year cannot be judged on that year alone.

FAQ

What is a reasonable maximum drawdown for a retail EA?
Many educational sources suggest 20-30% is towards the upper end of what most retail traders can tolerate emotionally, but the acceptable figure depends on the trader.

Does low drawdown mean low risk?
Not necessarily. Low reported drawdown can hide large open floating losses or a very short sample.

Is a 5% drawdown EA safer than a 30% drawdown EA?
Not automatically. You must know the leverage, the sample length, and the recovery pattern.

Can drawdown ever be zero?
Only on a very short sample. Any long sample will show some drawdown.

How much backtest data do I need to trust a drawdown figure?
At minimum several years, ideally covering different volatility regimes and central bank cycles.

Should I average an EA’s drawdown across multiple pairs?
Only if the EA genuinely trades multiple pairs with equal weight. Otherwise, evaluate per-pair.

Trading foreign exchange and CFDs on margin carries a high level of risk and may not be suitable for all investors. This article is for educational purposes only and does not constitute financial advice.

See EAs Evaluated on Real Risk Metrics

Drawdown is only one of several checks we apply before an Expert Advisor passes our review process. Explore our editorial view on vetted Expert Advisors to see how we assess risk, transparency, and long-term consistency.