Always Check the Worst Month in Your Backtest

Bottom line: decide by the worst month, not the average

When we evaluate an EA, the first thing we open is not the annual return or cumulative profit but the monthly results table. We find the worst month in it and ask: "Could I keep running this EA if this month happened three times in a row?" If the answer is no, we either skip the EA, no matter how good its results look, or lower the lot size before using it.

The reason is simple. Most people who quit an EA don't quit because of the maximum drawdown figure. They quit because of how it feels to see "I lost this much this month." Maximum drawdown is an event that comes once every few years, but bad months show up almost every year.

Why look at months?

Maximum drawdown is hard to feel

Maximum drawdown is an important metric, but it doesn't tell you how many months it took to dig that trough. Being worn down slowly over six months and falling all at once in a single month do very different psychological damage. Monthly results show you how fast losses arrive.

Bad months come in clusters

Markets move in phases. With a trend-following EA, losses tend to pile up when the market has no direction for a long stretch. For example, DONCHIAN ENGINE recorded a maximum losing streak of 11 in its backtest. If a losing streak runs across a month boundary, the bad month won't be a one-off. That's why we don't treat the worst month as a single event and instead assume it can repeat back to back.

Why multiply by three

There is no statistical basis for the number 3. It's our own way of building in a margin, based on the idea that "the worst in the past is not the worst in the future." Market conditions that didn't occur during the backtest period may still occur from now on. So instead of planning for the past worst month as it is, we allow for at least several times that.

Steps to check

1. Line up the monthly returns

Export the monthly profit and loss from the MT5 backtest report and look at it in percent, not in currency. An EA that compounds produces larger amounts later in the test, so comparing in currency makes the early losses look smaller than they were.

2. Pick out the worst month, plus the second and third worst

Look at the bottom three, not just the single worst month. If one month stands out as far worse than the rest, check what happened in it (an economic release, a sudden spike in volatility, wider spreads, and so on).

3. Count the consecutive losing months

Count the longest run of negative months. An EA with a long run is mentally harder to hold, even if its worst-month figure is small.

4. Apply "worst month × 3" to your own capital

As a general example, if the worst month is -10%, three times that is -30%. If your account lost that much, could you still avoid any impact on your living expenses and keep the EA running? If the answer is no, lower the lot size.

The last row is a supplement. If the effective DD is much deeper than worst month × 3, that EA has had periods when it kept sliding for many months. For example, ATLAS has an effective DD of 26.3%, and MEGAMAX DONCHIAN has an effective DD of 58.6%. Don't stop at the monthly numbers and feel reassured; cross-check them against the DD.

Limits and cautions

This method has weaknesses too, and we want to state them.

• Month boundaries are artificial. A large loss that spans the end of one month and the start of the next gets split across two months and looks smaller.
• A backtest is a replay of the past. Spreads, execution delays, and price differences between brokers will make actual monthly results differ.
• It is perfectly normal for forward results to be worse than the backtest. In the demo-account forward test we publish, ATLAS PORTFOLIO shows a return of -18.74% and a DD of 18.74%. These are figures from an Exness demo account as of October 23, 2026, not results from a live account. It has only just started, so we can't judge it good or bad from this result. We include it anyway as an example that periods worse than expected do come along.
• The method described here cannot prevent losses. It is only an input for deciding whether to keep going, and it does not guarantee profit.

You can compare each EA's metrics side by side in the measured rankings. For the thinking behind capital allocation, also read EA & MT5 know-how. If you want to compare on the same USDJPY H1, the DONCHIAN ENGINE and ATLAS pages include backtest details.

Summary

• Decide whether to adopt an EA by its worst month, not by the average or the cumulative result
• Look at monthly results in percent, not currency, and also check the bottom three months and the number of consecutive negative months
• Adjust your capital and lot size so you can keep running the EA even if "worst month × 3" arrives
• If the effective DD is far deeper than worst month × 3, suspect a long period of continued decline
• The worst month in a backtest is not necessarily the worst that is still to come
• Three times is a rule of thumb, not a number with a basis, so leave a generous margin