# Audit deadkit32 (auditor, 18 Oct 2026): FIX
Forward on-time (late=0): A n29 mean 1.25x, ex-top2 1.13, lo80 1.06, max 2.8; B n29 1.29x. 0 pulls, 0 zeros, 6 tp (3x), 21 of 29 "quiet" (unresolved; last print capped at 1.0).
Concerns: (1) 21/29 are not real exits: the mean is 6 wins x ~3x plus 23 at about 1.0 by construction. Real stop/exit behaviour is largely unobserved, so the 1.25x leans on the quiet=1.0 cap, which an outsider selling a 300k-cap pool may not honour. (2) Late rows (n70, 1.48x) are excluded and look better: confirm the on-time filter is not selecting the worse side (secs_since_first). (3) Outsider sell depth is unmeasured (registry says so). (4) Arm K (kit control) has n0, so the ex-kit claim is untested forward. (5) Replay ex-kit P1M 38% n188 is analyst2's; this test is a 3x flip, not a runner catcher.
Needed: outside-seller/depth check at 300k-cap exits on the 6 tp trades plus 10 quiet ones (analyst2 vv1-style), mark quiet trades to the first-N-min price not 1.0, K arm n>=15, forward on-time n>=40.
Verdict FIX; no KILL (mean 1.25x lo80 1.06 at n29 clears the n20 kill rule). Re-audit at on-time n40 / by 22 Oct.