May 2026 closed at -2.54% on our live breakout account. It is the first losing month since the opening weeks of the account (November 2025 closed -5.10% while positions were being established) in mid-November 2025, and instead of hiding it, this post walks through it, because how a system behaves in losing months is the entire reason to trade systematically.
The month in numbers
After five consecutive positive months (December +12.46%, January +2.17%, February +10.09%, March +7.21%, April +1.80%), May produced a string of failed breakouts. The account gave back 2.63%, with the drawdown from the April peak reaching roughly 2.8% intra-month. Win rate for the month dropped below the long-run 57%, which is exactly what a clustering of false breaks looks like.
Is this normal? The backtest says yes, emphatically
Across 11 backtest years, the breakout system had 34 losing months out of 127, roughly one in four. The average losing month in backtest is around -2.5% at this risk level. May 2026 at -2.54% is not an anomaly. It is the median bad month, arriving on schedule. If anything, five consecutive positive months without one was the statistically unusual part.
What we changed in response
Nothing. Not the risk setting, not the pairs, not the parameters. This is the entire point of systematic trading and the hardest part to internalize: a system with an edge produces losing months as a routine cost of doing business. Intervening after them (reducing risk at the bottom, pausing the system, tweaking parameters to fix the last month) is how traders convert normal drawdowns into permanent underperformance. The rules were written calmly years ago precisely so nobody has to make decisions during the uncomfortable weeks.
What we watch instead
We do not watch monthly P&L for signals. We watch structural statistics for drift: is the live win rate over a large sample still tracking the backtest? (Yes: 52.8% live vs 57.2% backtest, inside normal variance for this sample size.) Is the average loss per trade within its designed bounds? (Yes.) Is the drawdown within the historical envelope? (Comfortably: -2.8% against a backtest max near -14% at equivalent risk.) As long as the structure holds, monthly results are weather, not climate.
The full month, trade by trade, is visible on the MyFXBook verified track record. Losing months stay published forever. That is the policy.
What the losing trades actually looked like
Aggregates hide texture, so here is the texture of May's trades. The month's losses were dominated by failed range breaks on GBPJPY and USDJPY: the classic pattern where London resolves the Tokyo range, the position triggers, and New York fails to confirm, reversing the pair back through the range and out the stop. Individually, each was an unremarkable, rule-perfect trade losing its designed 1R. What made May a losing month was purely their clustering: several failures arriving without the usual interleaving of continuation days, because the yen pairs spent the month digesting rather than trending. No single trade was larger than designed, no stop was moved, no loss exceeded its budget. A losing month in a fixed-risk system is not an event. It is an arrangement of ordinary events, and recognizing that distinction is most of what separates traders who survive drawdowns from traders who narrate them into catastrophes.
The intervention urge, and the two sentences that answer it
Around the third losing trade of a cluster, a specific thought arrives on schedule: maybe pause it until conditions improve. It sounds prudent. It is mathematically incoherent, for two reasons that fit in two sentences. First, if conditions improving were detectable in advance, that detection would already be a rule in the system, and it is not, because in eleven years of data no such filter survived testing without destroying more profit than it saved. Second, the system's edge is the sum of all its trades, and the recovery trades that follow losing clusters are statistically indistinguishable in advance from the losing trades themselves: pause through the losses and you pause through the recovery, converting variance into realized underperformance. Every backtested version of discretionary pausing we have run lands in the same place: worse returns, similar drawdowns, plus a new dependency on the very human judgment the system exists to replace. The urge never fully disappears with experience. What experience builds is the reflex to notice it, name it, and let it pass unacted upon.