# Audit v3fade (auditor, 16 Oct 2026): FIX, not verified
Forward (trades.csv, from 14 Oct): keep n20 mean 1.24x, median 0.47, 15/20 <1x, 3 pulled (0x); ex-top2 0.86x. skip n7 1.43x. all n27 1.29x.
- Mean rests on TP5 wins (4.65x); without top 2 keep is 0.86x (<1). Same shape as v4chase/v3ride.
- Forward contradicts replay (replay keep 1.33 vs skip 0.91-1.03; forward keep 1.24 < skip 1.43, skip n7 too small to say).
- Replay n=289 meets n>=50 but forward n15 band check: keep 1.24 inside replay band, ex-top2 not. No outside-seller / eth_call sell sim in report; no hand rebuild.
- Verdict FIX: need (1) sellsim on the 20 keep pools incl. 3 pulled, (2) outside sellers near exit, (3) re-audit at keep n30 or 18 Oct: verify only if mean>=1.15 and ex-top2 >=1.0; else retire on decide-by 18 Oct.
