"Never Had a Losing Month" Over 24 Months Still Allows 1 in 8

19 November 2026, 10:20
Michael Thomas Harty
0
11

An EA with 24 months of history and not one losing month sounds safe. Statistically, it could still be an EA that loses one month in eight.

The idea: zero events in a sample doesn't mean zero risk. It means the risk is below some ceiling, and the ceiling is higher than most people expect. The quick version is the rule of three: with 0 events in n tries, the 95% upper bound on the true rate is about 3 / n.

Step 1. Apply it. 24 months, 0 losing months: 3 / 24 = 12.5%, or 1 month in 8. The exact figure is 1 − 0.05^(1/24) = 11.7%, so the shortcut is close.

Step 2. Check it the other way round. Suppose the true rate really is 12.5% a month. The chance of 24 clean months in a row is 0.875^24 = 4.1%. Rare, but not impossible, and with thousands of EAs on the Market, somebody's will have done it by luck.

Clean months seen 95% ceiling on the true losing-month rate
6 50% (3 / 6; exact 39%)
12 25% (exact 22%)
24 12.5% (exact 11.7%)
36 8.3% (exact 8.1%)

The part people miss: the shortcut flatters short records. At 6 clean months it says 50%, but the exact bound is 39%, so use the exact formula (1 − 0.05^(1/n)) below about 10 months. Either way, a perfect record needs years, not months, before it says much.

Where a tool fits: size the account for the drawdown the EA's trades imply, not for the months it has been lucky. VRD Quant resamples an EA's own closed trades 2,000 times and gives a p95 drawdown and a suggested operating capital. It is an estimate from the history you have, so a live drawdown can exceed it.

What to do today: take the longest clean streak you've been proud of, divide 3 by its length, and size the EA's capital as if that loss rate were real.

Try it: Try the free demo of VRD Quant in the Strategy Tester to see p95 drawdown and suggested capital on its labelled sample data, then run it on your own history. $69 to buy, or rent for $30 for 1 month.
https://www.mql5.com/en/market/product/204796

Virtual Risk Desk

Simplified, illustrative statistics: assumes independent months and a fixed underlying rate. Educational content, not trading advice.