# Analyst 2 idle run 48 (10 Oct 2026): 10 iterations, scripts work/analyst2/it48/

Pool types in focus (>=6 of 10): v4 no-hook native / WETH / USDG, hooked fee100 WETH. Data: radar/pool_paths.jsonl, r6nat/r6usd + v4sender, txfrom.db (tx.from, 4156 pools), sellprobe.db (new), crew_wallets.json, names.db, createtx.db.

| # | Lens | Result | Verdict |
|---|---|---|---|
| 1 | Exit depth (NEW data: sellprobe.db, 91 pools, 2 h) | Of 31 no-hook v4 pools shown at cap >= $1M, only ~14 let a $100 sell out at <10% impact (native 10/16, USDG 9/13 incl. dust reserves, WETH 3/5); the rest have reserve 0 or impact 100%. Real depth, not cap, decides whether a "runner" is sellable. | INFO; 2 h only, rerun when 3 d accumulate |
| 2 | Wallet: creator buys own token | 41/319 vs 16/346 reach 1M, but all of it is the buy in the create block = create-tx value (known). Creator re-buy 2-40 blocks later: 1/177, yet those pools never reach the band (cap5 median 16k). | DISCARD |
| 3 | Order-flow: two-sided flow, native no-hook born<100k (1626 pools, 27 to 1M) | sell/buy volume >0.3: 25/491 vs 2/569; zero sells: 0/160. Winners are 1-5-router pools with 100-200 swaps in 5 min and identical cap_first 26.9k = one kit batch (wash). | DISCARD (wash-farm marker) |
| 4 | Trap check: v4sender "sender" is the router (each address in 1000s of pools). My "distinct senders >=8: 0/22 native, 2/103 USDG" is a router count, NOT organic buyers. | Do not cite as organic breadth. Needs tx.from for native/USDG (request below). | TRAP / REQUEST |
| 5 | Wallets with tx.from, v4 WETH no-hook (1607 pools) | 35 pools reach 1M, 0 reach 3M, all army-heavy. Ceiling of the tier again. | DISCARD (dup of tier ceiling) |
| 6 | Meta: theme saturation (same name stem launched in prior 6 h, 13k tokens) | v3 P1M by prior count 0 / 1-2 / 3-9 / 10+: 3.5 / 4.5 / 3.7 / 2.9 %. Same flat on native, WETH. Being early in a theme does not matter here. | DISCARD |
| 7 | NEW lens: crew army in the first 5 swaps (tx.from in crew_wallets.json army/distributor/creator sets) | Crew-created hooked WETH pools: army>=30% of first 5 buyers 0/316 reach 1M; clean first 5: 77/258 (30%). Hooked fee100 only: 0/335 vs 5/183. v3: 0/29 vs 50/317. Splits early/late: 1/44 vs 17/138 and 0/25 vs 32/139 (v3). Tape-only proxy (median pool-count of the first-40 wallets) gives the same split, so it is not just the list. | MATURE-thin, INFO. Not an edge for current tests: the live band (cap5 250k-1M) has 0 army-heavy pools already; H100 trades tp3-4 and the proxy P3x is equal (.77 vs .84). Matters only for a 1M-tier pick on hooked fee100 (5 runners in 183). |
| 8 | Regime of the army | army-heavy share of new pools by day: 30-70% 21-29 Sep, 0-25% 30 Sep-7 Oct (sample drift partly). No pause rule. | INFO |
| 9 | Hooked organic (senders>=8) | router-count trap again (see 4), no use. | DISCARD |
| 10 | Request ppdata: tx.from + army label for native/USDG no-hook | below | REQUEST |

## Caveats
- crew_wallets.json is rebuilt continuously (mtime 10 Oct 13:00): the list could contain wallets added because of those pools. The tape-only proxy replicates the split, but a dated snapshot is needed to call it out-of-sample.
- txfrom.db covers band + control + hooked groups only, so absolute rates are enriched; contrasts within a group are fair.
- pool_paths max_cap multiples are inflated (hooked P2x 93%); I only used 1M reach.

## Request (ppdata)
tx.from of every swap in the first 5 minutes for all no-hook v4 pools with venue native and USDG born 13 Sep-10 Oct (r6nat 1626 + r6usd 6637 born<100k pools; about 0.5-1M swaps), into txfrom.db (same schema: pool, block, li, tx_from), plus a dated copy of crew_wallets.json from 6 Oct (any snapshot older than the first pools in the test). Test: distinct non-crew tx.from buyers in first 5 min vs reach 1M/3M, with the dumb baseline (same venue, same exit), split before/after 6 Oct. Pass: organic buyers >= 8 lifts P1M above 2x baseline in both halves with sell-probe reserve >= $5k. Fail: the unwatched native/USDG runners are crew prints, close the blind-spot rows.