# mc5 exit rule audit (analyst1, 18 Oct 2026, board #3077 direction 2)
Data: 101 closed live mc5 trades (fm_mirror) x v3live swap path 4h after buy. Script: /opt/crew/work/analyst1/mcx.py.
Rug pricing: price <=0.1x while no stop is active => exit 0; trades whose live exit was <0.1x (19 of 101) forced to 0 in every no-stop window. Fills at swap price, no exit slippage (upper bound on big winners).
Holdout: trades before 12 Oct (n=56) vs from 12 Oct (n=45).

| rule | all mean / ex-top3 | <12 Oct | >=12 Oct |
|---|---|---|---|
| live actual | 1.50 / 0.81 | | |
| base stop 0.5 | 1.71 / 0.98 | 1.86 / 0.73 | 1.51 / 0.65 |
| stop 0.3 | 1.86 / 1.14 | 1.80 / 0.67 | 1.93 / 0.84 |
| no stop 4h (rugs=0) | 2.48 / 1.69 | 2.66 / 1.29 | 2.25 / 1.09 |
| no stop 1h, then stop 0.5 | 2.21 / 1.49 | 2.15 / 0.90 | 2.29 / 1.12 |
| no stop 1h, then stop 0.3 | 2.24 / 1.52 | 2.21 / 0.97 | 2.27 / 1.11 |

Verdict: a grace window (no stop for first 60 min, then 0.3-0.5) beats the base on mean in BOTH halves (+0.3 to +0.8) and on ex-top3 in the later half (1.1 vs 0.65). Pure no-stop is best but is the most exposed to unsellable rugs/slippage; 1h grace keeps most of the gain (~90%) and the loss on rugs is already priced at 0. Early stops are what kill the 20-50x runners (16 stopped at 0.5x then ran 5-52x, mostly in first hour).
Not verified: exit slippage on 20x+ sells, pulled pools where swaps simply stop (covered only through live-rug override), n=101 with ~38 runners.
Proposal: shadow/paper arm of mc5 "stop disabled for 60 min after buy, then 0.4"; judge on fwd n>=15. Live change is Ben/exec only.