When fixed stops become noise: the −90% postscript

After the H1 swing system's rise and fall (E-07), one variant remained on the board — the one every trading book, every course, and every funded-account program insists on: replace the ATR-scaled exits with disciplined fixed-risk management. Fixed stop distances. Uniform risk per trade. The advice is so universal it reads as hygiene rather than hypothesis. We ran it through the same walk-forward harness as everything else.

The account lost roughly 90%.

Not a degradation — a demolition. The same signal, the same folds, the same costs that had (deceptively, per E-07) produced +$8,877 with volatility-scaled exits was nearly wiped out by the “safer” exit scheme. The gap between those two outcomes is the entire content of this postscript.

The sample this system traded spanned gold at $1,805 and gold at $5,596 — including calm stretches where the hourly bars barely breathed and the late-2025 parabola where hourly ranges exploded. Volatility was not a constant; it was the single most variable thing in the dataset. A fixed stop calibrated to look sensible on average is, in the loud regime, narrower than the market's ordinary hourly wiggle — inside the noise floor. Positions were stopped out by fluctuations that meant nothing, over and over, before any thesis had time to resolve.

Fixed-risk sizing then compounds the damage: a stop that is tight relative to current volatility mandates a larger position to hold risk-per-trade constant — so each meaningless noise-stop extracts a full risk unit. The discipline is intact; every loss is exactly the planned size. There are simply far too many of them. Death by a thousand correctly-sized cuts.

Exit geometry is not a style preference — it is an implicit claim about the distribution of noise. ATR-scaled exits claim “noise scales with recent volatility” (approximately true, which is why E-07's exits at least let the signal express itself). Fixed exits claim “noise is constant” — measurably false in any market whose price triples. When the claim is false, the stop stops being a boundary on your losses and becomes a metronome for them. “Risk management” that ignores the volatility regime does not manage risk; it formalizes it.

This was the last experiment in the program. The final scorecard: eight strategies across order-book microstructure, scalping, indicators, breakouts, mean reversion, and swing trend-following — every one dead under a single, unmoved honesty bar; one instrument validated (the volatility-expansion classifier that now drives the live regime reading); and a stopping rule (E-06) honored as written. The trading-research thread is closed. What replaced it is this site: the measurements published, the failures documented, the methodology public — because the graveyard turned out to be the most valuable thing we built. How we work, in full: the methodology.