How to Turn a Backtest Into a Monthly Expectation, and Why We Don't Publish One

Buyers ask us for a monthly percentage more than for any other number, and we do not publish one. Not because it is hard to compute — it is one division — but because on its own it is the most misleading number a backtest can produce. Here is how to compute it from any report, and what to put next to it so it stops lying.

Divide the net profit over the whole tested period by the number of months. A 36-month test that ends 108% ahead averages 3% a month. As an average that is not wrong. What it hides is how the months were arranged, and the arrangement is the only thing you will actually experience.

Open the monthly breakdown and count: how many months were negative, what the worst month was, and how many months produced most of the profit. With a typical trend-following system, a third or more of the months lose, and the average is carried by a handful of strong ones. Continue the example: 14 of 36 months negative, worst −9%, five months delivering most of the total. That system genuinely averages +3% a month, and it also genuinely sat through six flat or negative months in a row. Both are true of the same report; only one gets published.

If the EA sizes by percent of balance, results compound, and what you make in a month depends on the balance you brought into it. Whoever started before the strong stretch made most of the gain; whoever started right after it spent half a year underwater with identical settings. Same average, different outcome. A fixed-lot test scales differently: each month's gain is a fixed amount in currency, so the percentage shrinks as the account grows. Comparing the two by monthly percent is comparing two different objects.

A seller who quotes a monthly percentage without saying how many months lost is quoting the mean of a skewed distribution and letting you imagine it as the typical month. The typical month of most trend systems is close to zero; the mean is pulled up by the tail. That is not fraud, it is what an average of an uneven series looks like — but it is why the number is worth little alone.

Average month: net profit divided by months. Share of losing months: losing months divided by total months. Worst month: the lowest single monthly result. Every tester report contains the data for all three, and a seller can produce them in five minutes; if they will not, that tells you something too. Side by side — +3%, 39% of months negative, worst −9% — the same backtest reads very differently from "+3% monthly", and much more like what your account will do.

Nobody stops an EA during its average month. People stop during the fourth or fifth losing month, usually just before the stretch that carried the average. So the question is not "how much does it make on average" but "what does its typical bad half-year look like, and would I still be running it at the end of it". If the answer is no, the average is irrelevant, because you will never be there to collect it.

Our three best-selling EAs: GOLD NEURON (AI) https://www.mql5.com/en/market/product/187329 · ATLAS PORTFOLIO https://www.mql5.com/en/market/product/182751