# Miss review 2026-10-19 (analyst1)
Source review/misses.csv, born 19 Oct, deduped by token: 49 tokens = 20 real runners (ath 1M-20M) + 29 phantoms (ath 2.4T+, bad price math; the "cap $2.4e6M" strings and the 20M+ x30 count are this unit bug, not real). Real tiers: 20M x1 (UHI), 10M+ x4, 5M x4, rest 1-3M.

| group | n | status | cause | would catch |
|---|---|---|---|---|
| v4:native organic (cap_first 2.2k-9.7k) | 13 | mc5 NOT SEEN (venue pool_uniswap_v4); aipaper never scored | venue not watched. Repeat kits: TRENCHBOT x3, RAETO x3, FTR, PIERRE, COMPUTE (serial relaunch). Includes UHI 20M, 0x000050 14M | v4:native into mc5/aipaper radar; entry at first low cap |
| v3 thin (cap_first ~800) | 3 | mc5 MISSED, cap under band | 0xa1e4eb 11M, 0x149391 11M, 0xea26dc 7M; thin seeds, high pull risk | thin arm, needs pull sim |
| v3 jump / organic | 3 | MISSED, rule | one-offs | none |
| v4 hook weth | 1 | SNP 10M, seen, not bought (cap 241k, over band) | rule | none |

Technical failures: 0. No paper test caught any of the 20.
Same picture as 18 Oct (miss-2026-10-18.md): v4:native organic is the biggest block (13 of 20; 2 of the top 5 tiers) and repeats daily. Existing ledger entries cover it (v4:native feed/arm, v3 thin); caution stands: replay needs pull/outside-sell sim (#651/#654, kit pools pulled ~99.8%). Phantom filter (ath < 2e9) should be applied in review.py so the tier counts are honest.
## One change
Finish the v4:native watch (ppfeeds miss-watch) so mc5/aipaper score those pools; then paper arm first-cross entry at cap <=30k with pull sim. No new idea beyond ledger.