The clustering problem
This explainer uses an illustrative dataset of 1.2M anonymised stop-loss placements on EURUSD and XAUUSD over a single quarter. The result is a map of crowd behavior: stops pile up exactly where every textbook says to put them — just beyond round numbers and the most recent swing high/low.
Where retail stops sit
% of 1.2M stop orders
Clustered stops are visible liquidity. Price is repeatedly drawn to these pools — the “stop hunt” traders complain about is mostly the market mechanically filling where the resting orders are.
What clustering costs
Comparing outcome distributions, structure-hugging stops were tagged-then-reversed (stopped out within 10 pips of the eventual turn) 2.3× more often than ATR-scaled stops on the same entries. The fix is mechanical:
- Scale stops by current volatility (e.g. 1.5 × ATR-14), not by chart structure alone.
- If the logical level is 1.1000, the wrong stop is 1.0995 — and so is everyone else’s.
- Wider, volatility-aware stops with smaller size carry identical dollar risk and dramatically better survival.
Educational content only. Figures, benchmarks and charts are illustrative examples, not IUX production data or a representation of trading outcomes. Nothing here is investment advice.