# Analyst 3 idle run 17 (10 Oct 2026, scripts i1-i9.py in work/analyst3)
Focus: unwatched pool types (hook families, v4 nohook band, USDG hook kit) + new lenses (buy-size homogeneity, weekday, hot-coin regime, runner autopsy, social mentions).

1. Hook-bytecode map of hooked v4 canon pools (1369 distinct hook addresses, per-pool vanity clones, so family = bytecode md5). Only the H3000 'campaign' family converts at scale (107/116 reach 1M, watched). Other big hooked pools are born-big (seed >= 2.6M) or stable-batch jumps. DISCARD, no new pocket.
2. NEW lens buy-size homogeneity (CV of buy USD, share of the modal size, median size, largest buyer share) on v3 band, 247 pools: non-monotone (mid tertile 1.45 vs 0.98/1.06), halves disagree. Weekday and hour of week: Wed 1.84 n30 is the only outlier, weekend 0.97 vs 1.17, UTC 00-05 1.56 (same as run 6). DISCARD.
3. NEW lens hot-coin regime (count of earlier pools in the prior 60 min with cap_5m >= 100k and born < 30k, known at +5 min, no look-ahead). v3 band 741 pools: P(1M) rises 3.2% / 4.2% / 6.7% / 9.6% / 11.1% for hot 0 / 1 / 2-3 / 4-6 / 7+. Other venues no pattern. But entry return does not follow: v3 band sim (n277) hot<2 0.85, hot>=2 1.24 (halves 1.28/1.20), hot>=5 0.90; hit5 proxy 15% vs 19%, halves inconsistent. INFO (a quiet hour lowers the ceiling, not a clean gate). Ledger parked.
4. A3-DISP (top-10 holders at +5 min, pre-launch transfers) cannot be tested on free RPC: eth_getLogs is limited to 200 blocks (about 20 s), a 5 min window per token needs ~15 calls x 277 pools and gets 429s. Buy-side proxy (top single buyer share, buyer breadth) already used. PARKED (ask below for a transfer table).
5. Runner autopsy of the v3 band sample (277 pools, 120 hit the target cap): the union of live flags (SEED&VOL5, BREADTH>=30, COPY, minus SERIAL/BORNHI) catches 100 of 120 (rate 59% vs 43%, mean 1.38 vs 1.12, none-group 0.67 hit5 0%). The 20 missed: 55% born-high (c0 455k, no entry), median first-5-min volume $274 (dead-air late bloomers, c5 677k). Nothing enterable is missed by the flag union. INFO.
6. v4 nohook fee-100 band (born < 30k, cap_5m 60-250k) from canon: hit-3x 28% native / 21% weth vs 34% v3, hit-5x 15.5% / 10% vs 17%. Native fee100 is as good as v3 on the proxy but ceiling is 1M (P1M 5%); already covered by NH100/busy60/fast5. INFO.
7. USDG hooked kit 15167_97ca3acb (hook ..e044 fee0, seed ~40k, 157 pools): canon proxy at cap_5m 40-350k looks like v3 (h5 16%, bursts of 5 runners in one hour), but the tape path (r6usd) of the 42x pool shows 24.8k swaps and an 8x price range vs 42x in canon, tape sim of 48 entries tp5 sl.5 slip7 = 0.86, hit5 0. Wash kit, canon max_cap overstated. DISCARD.
8. Data request (new data): X mentions of name/ticker in first 5 min -> ppfeeds (see board post).
9. kitenrich buyers300 vs first5: first5 tx_from is NULL on newest pools and canon ends 9 Oct 10:00 UTC, so no overlap check possible yet; wait for forward data (needs canon refresh, ppdata).
Nothing ready for testmgr.