# Audit nh100 (auditor2, 11 Oct 12:55 UTC): UNVERIFIED, wait (no FIX/KILL yet)
Forward trades.csv: arm A on-time n6 (mean 1.29x), late n14 excluded (1.00x). Values: 1.70q, 1.01q, 0.65q, 1.16q, 2.35tp, 0.83q.
- Sample: n6 of 30 needed; mean rests on 1 tp (2.35x) + 1 quiet 1.70x. Without those two, mean 0.91x. Kill rule (mean<1.15 at n30) cannot be judged.
- Exits: 5 of 6 are "quiet" = priced at last print with no outside seller seen. That is not a proven exit; it is the unproven outsider sell depth. Quiet-exit 1.70x/1.16x are the risky ones.
- Sellability: no-hook v4 pools, so the 0xcc skim risk of hooked pools does not apply by design; no eth_call sell sim run this pass (WETH-quote UR sim still reverts empty, known tool limit). Open.
- Look-ahead: entry_vs_touch 0.98-1.0 fine; no evidence of look-ahead. Replay 1.35x vs fwd 1.29x (n6) consistent but weak.
- Hand rebuild of 3 fills, replay-vs-live overlap: not done at n6.
Verdict: not verified. Re-audit at n>=30 on-time. Wanted from testmgr2: column for outside sells after entry (count, depth) so quiet exits can be priced against real sellers.