Two weeks ago we published a post about May closing at -2.54%. June is currently also running negative. Two consecutive losing months is where most EA buyers turn their systems off, so let me use our own live account as a real-time case study in reading a drawdown correctly.
The situation as of mid-June
The account peaked in late April around $131,000. Through May and the first half of June, a series of failed breakouts has pulled the balance down several percent from that peak. Individual losing trades in the -$1,200 to -$1,500 range have clustered, which at this account size is the designed roughly 1% risk per trade doing exactly what it says.
The question that matters: is this inside the envelope?
Every system has a historical drawdown envelope. Ours, from 11 years of backtest at equivalent risk: max drawdown near -14%, with multi-month losing sequences appearing in 2016, 2017, 2019, and 2024. Consecutive losing months occurred repeatedly in the backtest and every single one was followed by recovery to new highs, which is definitionally true of a backtest that ends at its peak, but the depth statistics are the point: the current live drawdown of roughly 5-6% from peak is barely a third of the historical maximum. This is a Tuesday, not an emergency.
The math of when to actually worry
A rational abandonment rule should be written before the drawdown, and it should be structural, not emotional. Ours looks like this:
- Depth trigger: live drawdown exceeding 1.5x the backtest maximum would indicate the live market differs structurally from the tested one. For us that means roughly -20% at current risk. We are nowhere near it.
- Statistical drift trigger: live win rate over a rolling 100+ trades falling more than 10 points below backtest. Current live win rate: 52.8% vs 57.2% backtest, inside the variance band. No structural drift.
- Mechanical trigger: the system taking trades outside its own rules, indicating a code or broker issue. Zero occurrences.
What discipline actually looks like
It looks boring. The system keeps trading, the results keep publishing to MyFXBook, and no parameter gets touched. If the drawdown deepens, these posts will keep documenting it, because a track record that only narrates the good months is marketing, not a track record.
The information in maximum adverse excursion
One diagnostic deserves a deeper look during any drawdown: maximum adverse excursion, or how far trades went against the position before resolving. In a healthy drawdown, losing trades hit their stops at the designed distance and winning trades show normal heat: the system is losing because outcomes clustered badly, not because trades are behaving strangely. In a structural breakdown, the texture changes first: winners start showing unusually deep adverse excursions before recovering, stops start filling with abnormal slippage, the average time-in-trade drifts. Through May and June, our trade-level data shows designed behavior throughout: stops filling within normal slippage tolerances, winning trades showing historically typical heat, holding periods inside their usual distribution. The account is losing money exactly the way the backtest said it sometimes would, which sounds like cold comfort but is actually the entire diagnosis: the machine is running to specification through an unfavorable stretch of road.
What we would do if a trigger actually fired
Pre-committed triggers only discipline behavior if the response to them is also pre-committed, so for completeness, here is ours. A depth breach (drawdown exceeding 1.5x backtest maximum) or a structural drift breach (rolling 100-trade win rate more than 10 points below backtest) does not mean turn everything off and it especially does not mean tweak parameters until recent results improve, which is just curve fitting with extra fear. The committed response is: reduce risk to 25% of normal (keeping the system live so data keeps accumulating), open a formal investigation comparing live trade signatures against backtest signatures to locate the divergence (execution, market structure, or code), and only resume full risk when the divergence is either explained as benign or corrected at its source. Turning a system off entirely destroys the very information needed to diagnose it. Keeping it alive at token size preserves the instrumentation while capping the cost of whatever is wrong. This protocol has never yet been triggered live. Its value is that it exists in writing before the day it might be, because protocols written during emergencies are just panic with formatting.