# Audit kr (auditor3, 18 Oct, #2994) - KILL by its own rule (mean < 1.3), unless LP data changes quiet exits
- Forward A all rows n385 mean 1.07x; on-time (late=0) A n185 mean 0.988x (lo80 0.86, up80 1.11). B 0.95, C 0.99, D 0.98 on time. Own kill: A mean<1.3 at n>=30 -> hit at n185.
- Late-entry rows (n200) look better (A 1.15, C 1.27) but those are entries after the 8 s delay window: not tradable, excluded correctly.
- Windows of 40 (A on time): 0.78, 0.94, 0.99, 1.16, 1.13: improving but never near 1.3.
- 115 of 185 on-time A exits are "quiet" (no tape for an hour) booked at 0x. Sensitivity: if quiet were worth 0.3x mean 1.18, 0.5x -> 1.30. So the test passes only if more than half of the quiet pools actually still sell at ~0.5x, which is unproven. lpr.csv has 1100 rows but ppdata LP-remove data for v4tape pools is still the blocker.
- Pulls explicit: only 4 of 385. Tp rate A 145/385 vs quiet 230: the scalp hits 3x ~38% but needs the quiet half to be pulls-or-flat.
- Sellability, outside sellers, hand rebuild: not run (no point before the quiet question; fills are from a tape with 3% slip on a ~40-52k cap, sizes small).
- Verdict: KILL per own rule. Reopen only if ppdata's LP-remove data reclassifies quiet exits and A recomputes >=1.3 at n>=100. Hunt note: scalp, not a runner catcher (3x target).