# QSR risk-ladder backtest: over-stopping check — 2026-08-02

15 real closed qsr trades (data/trades.jsonl), each trade's actual worst price during its hold (entryPrice × (1+maePct/100)) compared against a hypothetical tightened stop price (signal.price × (1-capPct/100)) at both risk-ladder tiers. Read-only, no orders, no re-fetch — maePct was already computed at close from real intraday bars.

## Per-trade data

| Symbol | Entry | P&L | MAE | Exit reason | Tier1 (6%) hit? | Tier2 (4%) hit? |
|---|---|---|---|---|---|---|
| TDG | 2026-07-22T13:33 | +4.35% | -1.56% | take_profit | no | no |
| NFLX | 2026-07-22T13:34 | +4.25% | -2.26% | take_profit | no | no |
| ISRG | 2026-07-27T14:18 | +4.86% | -0.33% | take_profit | no | no |
| NXPI | 2026-07-22T13:33 | -7.58% | -8.06% | stop_loss | **YES** | **YES** |
| NXPI | 2026-07-22T13:33 | -8.28% | -8.06% | stop_loss | **YES** | **YES** |
| ISRG | 2026-07-27T14:18 | +1.98% | -0.32% | trailing_stop | no | no |
| NFLX | 2026-07-22T13:33 | +2.84% | -2.34% | trailing_stop | no | no |
| UMC | 2026-07-29T19:49 | +4.04% | -1.67% | take_profit | no | no |
| UMC | 2026-07-29T19:34 | +4.09% | -3.13% | take_profit | no | no |
| UMC | 2026-07-29T19:34 | +10.69% | -3.13% | trailing_stop | no | no |
| UMC | 2026-07-29T19:49 | +12.37% | -1.67% | trailing_stop | no | no |
| CARR | 2026-07-30T15:02 | +4.02% | -1.17% | take_profit | no | no |
| CARR | 2026-07-30T14:47 | +3.86% | -1.94% | take_profit | no | no |
| NXPI | 2026-07-31T13:40 | -7.41% | -7.44% | stop_loss | **YES** | **YES** |
| NXPI | 2026-07-31T13:40 | -7.41% | -7.44% | stop_loss | **YES** | **YES** |

## Result

0 of 11 real winners would have been stopped out early at the 6% (tier 1) cap.

0 of 11 real winners would have been stopped out early at the 4% (tier 2) cap.

**No real winner's price path breached either tightened cap before its actual exit** — the caps look safe against this sample.

> Caveat: the risk ladder only ever engages for a symbol with a recent QSR loss (strategy/qsr.ts's qsrRiskLadderTier). None of the winners in this sample had a prior QSR loss at entry time, so the ladder would never actually have fired for them historically — this check is a more conservative stress test of the cap VALUES themselves, not a claim the ladder changed any real outcome. Small sample (13 winning trade legs across 7 buy events as of this run) — re-run as more QSR volume closes, especially once a symbol genuinely re-buys after a loss.