Backtest entry-anchored stop-loss rules with entry cohorts, not signal-start simulations

A trading strategy was backtested as "hold whenever the trend signal is ON" (no stop), then deployed live with a fixed percent-from-entry stop-loss overlay added for safety. The continuous backtest showed the stop rarely firing, but live it fired twice in six days and realized a large whipsaw loss. Root cause: in a continuous simulation, the position's entry price dates back to the signal's start (often years earlier), so an entry-anchored stop is far out of the money and almost never triggers — the sim silently understates the stop's cost. A real account entering mid-trend has a fresh anchor, so the same stop behaves like a tight stop and churns.

Test entry-anchored rules (percent-from-entry stops, breakeven moves, trailing-from-entry logic) with an entry-cohort simulation: for EVERY day the signal is ON, start a position that day, run it forward over a fixed horizon (e.g. 6 months) under each rule variant including re-entry logic, and compare the full return distributions (5th/25th/median/75th/95th percentiles and mean), not point estimates. In this case cohorts revealed the stop cut median AND mean returns in both the full sample and the recent sample, left max drawdown unchanged, and on one asset made even the 5th-percentile tail WORSE (stop-out + re-entry churn compounds). The fix adopted: widen the stop to a catastrophe-only level scaled to the vehicle's leverage (a level the underlying only reaches on structural breaks), keep the signal itself as the real exit, and let account-level kill-switch rules carry tail protection. Also test the re-entry rule jointly with the stop — a "wait for N green closes" confirmation looked prudent but produced the worst tail of all variants.

An entry-anchored stop's behavior is path-dependent on WHEN the position was opened, so any simulation that only enters at signal transitions samples a tiny, unrepresentative subset of possible anchors. Overlays added after backtesting are effectively untested modifications; and on leveraged vehicles a "wide" percent stop maps to a routine move in the underlying (e.g. -15% on a 3x ETF is only a -5% index day), turning normal volatility into realized losses.

Any systematic trading or portfolio automation where a stop-loss, trailing stop, or other entry-price-anchored rule is layered onto a signal-based strategy — especially on leveraged ETFs or high-volatility assets, and especially when the live account enters mid-trend rather than at the signal's start.