After 14 months of live trading, I reviewed my records and found a recurring problem: inconsistent stop-loss execution. Sometimes I’d cut at 8%, other times I’d hold through a 20% drawdown. This wasn’t a discipline issue — the rules themselves were never clearly defined.
The problem with percentage stops
Traditional percentage-based stops (e.g., a fixed -8%) look simple, but they ignore a crucial factor: different stocks have vastly different volatility profiles. For a stock with 5% daily swings, an 8% stop barely gives it room to breathe. For a 1% volatility stock, 8% is far too loose.
The ATR approach
ATR (Average True Range) defines “abnormal” movement based on each stock’s own volatility. Using 2× ATR as the stop distance — if a stock’s 20-day ATR is 3%, the stop is -6%; if ATR is 1.5%, the stop is -3%.
I backtested this on CSI 300 constituents using Qlib + Backtrader. The conclusion was clear: ATR stops significantly outperform fixed percentages in controlling max drawdown, with less drag on returns.
Key insight: Good stop-losses aren’t about execution discipline — they’re about whether the rule matches the asset’s character. ATR turns stops from “one size fits all” to “tailored.”
Next step: combining ATR stops with a market regime classifier. Update coming next week.
