# a2-1447: missed runners as a group (misses.csv, 88 runners, born 7-10 Oct) + p204 kill follow-up
Analyst2, 11 Oct 2026. Source: review/misses.csv (305 rows = 88 unique tokens x bots).

## What the 88 missed runners share
- Venue: v4 no-hook is the biggest labelled group: 32 of 65 (WETH 21, USDG 7, ETH 4). v3 launcher 11, v3 no-hook 8, pons_v2 4, rest singles. 23 are day 7 Oct with venue "?" (pre-labelling, unusable for venue stats).
- Share of >=$5M runners: v4 no-hook 8 of 32; v3 launcher 6 of 11 (!); unknown 12 of 23.
- Seen vs not: mc5 NOT SEEN on all v4 no-hook (by design: it watches only the v4 ETH pool type it knows). fml2 saw v4 no-hook but MISSED all 21 (a score/entry decision, not a blind spot). aipaper5 NOT SEEN on 88 of 88: it scored nothing in 4 days (dead or unfed; one row per token = 100% blind, worth a look by testmgr).
- "never scored" (~26 tokens, all v3 launcher / v3 no-hook, e.g. VORTA $20M, swordcat $15M, SOCKET, DAEMON, hoodcat, PAWS): the AI pipeline has no feed for v3 non-fee100 pools. Biggest unwatched pool class for the mission (VORTA, hoodcat are 17-20M).
- Launcher kits: each launcher address (10 of them) produced ONE runner. No repeat creators => a launcher-address allowlist/blocklist is useless; the kit (bytecode hash) is the unit, not the address. Meta: stock/ETF-ticker names on USDG v4 (AAPL x2, QQQ, STANDARD, BONER, OpenAI) = 7 runners incl. AAPL $309M: a USDG-quote meta nobody watches; re-launches of the same name recur (AAPL, REALM, V4, LOL, PAWS, DAEMON, hoodcat, CHOP x5): same ticker re-launched after a first run.
- Timing: 4 runners were already > $1M at 5 min (SWARM, hoodcat 2.2M, SIB, BMBFI 2.6M): mc5 entry at +5 min is too late for them; they are first-seconds runners.
- Sizing: 0x73991a25 runners (ClapCat, roflmao) were $107-116k at 5 min and the family is sale-locked (audit p204): these two are NOT real misses, remove them from the miss list.

## Check of yesterday's proposals
No machine-readable proposal list in the review dir; the history.jsonl gates show fmmc5 flipped to "READY FOR BEN TO STEP SIZE" and fml2 0/2 days. Not re-litigated.

## p204 / #1450 (auditor3 kill)
Accepted. Idea 47 (v3 fee100 low-cap band = p204 universe) was already dominated (#111). Rule from now: any v3-launcher test requires a sell-gate (fresh-address transfer to pool must pass via eth_call) and excludes code hash 0a493d1af3d0 / launcher 0x73991a25. Applies to my open H3000 late-300k entry (#483): its 107 'P1M 92%' set includes launcher pools? To check before testmgr's fill sim; asked below.

## New test ideas (to ledger, none sent to testmgr: need data first)
1. Sell-gate filter as a universal pre-entry check (all venues): count how many "runners" in v3 launcher and pons_v2 are unsellable; the true miss list shrinks. Needs ppdata: eth_call sell test for the 88 tokens.
2. Same-ticker relaunch: when a name that ran before is re-launched (AAPL, REALM, LOL, CHOP), P(runner)? Needs tokmeta names (ppdata name backfill #422) vs pool_paths.
3. First-seconds lane for v4 no-hook + USDG: entry <60s at <100k for the 4 runners born >1M-at-5min.
4. aipaper5 audit: why 88/88 not seen.