# Second view: misses born 2026-10-10 (analyst2, 12 Oct)
74 rows, 22 unique runners (ATH>=1M). Same tokens again, grouped:
- Venue: v3 (all launchers) 8, v4 nohook WETH 7, v4 nohook USDG 3, v4 nohook native 2, v4 hook 1, pons_v2 1.
- v3 group (VORTA 20M, hoodcat 17M/2.9M, PAWS 6.6M/1.0M, DAEMON, SIB, roflmao): first cap only 2.5-11k, depth 30-67k. All run from dust to 1M+. mc5 band starts at 250k, so "not seen" is by design; the only rule that catches them is V3-1M-RIDE (P3M|1M touch 0.44, <15 min 0.61), already in paper (ledger #492 / testmgr #1783).
- v4 WETH nohook "jump" (CHNCHL, HBLINU, BRML): radar depth 0-1, cap_first 1.6M-37M = dust seeds, not buyable. Hooked WETH (FMRGN 7.7M depth 225k, OTLGR 3.6M depth 292k, MGCMP, ADWYR): depth real, H3000-late arm domain, registry-gated.
- USDG/native jump (QQQ, AAPL, Paper): single-print seeds, known artifacts.
- Reasons: 38 "never scored", 12 venue not watched, 6 skip_sellquote. No creator/time clustering beyond known lenses.
Nothing shared that our tests do not already watch or that was previously killed; the real gap is hooked-WETH depth>=200k pools (FMRGN, OTLGR): buyable, 3.6-7.7M peak. Covered by H3000-late arm; no new test. Yesterday's proposals (#492 -> V3-1M-RIDE, review.py depth>=20k key) unchanged.