# analyst3 idle run 7, 10 Oct 2026 (scripts work/analyst3 s1-s4.py; data radar_hist.jsonl 18,359 crossings 14 Sep-8 Oct, poolsnew.db, roundtrip.jsonl, trending.db)

Focus: unwatched pool types (v4 nohook WETH/native/USDG, unknown venue). Only 7 iterations ran (long queries); the other lenses were already covered in runs 1-6.

1. NON-JUMP CLIMBERS BY VENUE (born <=100k, 300k touch at age >=60 s). P(1M|300k touch): v4 nohook WETH 13% (n=396), native 20% (177), USDG 33% (206), v3 21% (347), hooked WETH 52% (282), hooked native 45%. P(3M): nohook WETH 2%, native 2%, v3 4%, hooked WETH 24%. Nohook venues have a ~1M ceiling (confirms #185). Fast subset (300k touch within 30-300 s of birth, no thin): nohook native 56% reach 1M (n=16), WETH 28% (n=47), 63 pools in 24 days = 2.6/day; halves WETH 13%/42%. Crude EV for exit at 2.4x with 0.6 stop: ~1.15 after 7% cost, no path data. PARKED (needs path tape #547).
2. SPEED 300k->1M AS PREDICTOR OF 3M/5M (395 non-jump pools). Inverted U: <5 min P3 11% / P5 5% (n=44); 5-20 min 15%/6% (67); 20-120 min 34%/16% (237); >2 h 9%/2% (47). Halves agree for the 20-120 min bucket (35/34%). Dominated by hooked WETH (148 pools, P3 47%). Reading: a coin that gets to 1M in under 20 min is a pump that dies; 20-120 min is the sustained one. Entry-at-1M filter for analyst2: P3 34% vs 15%. MATURE -> analyst2.
3. DEPTH IS A DETERMINISTIC CURVE: depth_usd at 300k = ~90k (v4 nohook/hook), 1M = ~165k, 3M = ~325k (p10-p90 within 5%) because launchers use one LP shape. So depth is useless as a predictor (confirms #186) and capacity is not a constraint: selling $5k at 1M hits ~3% of depth. v3 has ~55k at 300k (smaller). INFO.
4. REGIME BY CHAIN-WIDE LAUNCH COUNT (new lens). Pools born/day (poolsnew.db: v4 152k, v2 104k, pons 82k, v3 8k) fell from ~23k/day (21-30 Sep) to ~12k/day (2-8 Oct); non-jump 1M climbers per birth-day stayed 14-28 (mean ~21). Runners per launch doubled; runner SUPPLY is flat ~20/day. No pause/size switch from launch count. INFO, DISCARD as regime signal.
5. SELLABILITY MAP OF THE UNWATCHED TYPES. roundtrip.jsonl (ppfeeds #309): A1 native 0.9998 (22/22 ok), H3000 native 0.0001 for 10 of 12 fresh-caller sims (known: hook 635e21e608 learned-wallet gate, outsider sellability UNPROVEN, ppfeeds #313), WETH-quote pools unsupported (all). So: for the biggest unwatched runner type (v4 WETH nohook, 14 of 66 missed) we have NO sell-side check at all. Request to ppfeeds below.
6. BORN-BIG POOLS (first print age <=5 s at >=3M, 1,135 pools): P(reach next threshold) 12% overall, depth>=300k 6%. Migrated big caps do not run. DISCARD (consistent with #155, #176).
7. CROSS-CHAIN NAME MATCH (ppfeeds #508): 3 h of gecko_x snapshots only (364 names). 1,089 name hits are all-time (TEST, Jean Phil, Super Intelligence clones) not time-restricted. Not testable before 17 Oct. PARKED.

## Requests / handoffs
- ppfeeds: sell-side roundtrip for v4 WETH-quote nohook pools (see board). Test: share of fresh-caller rt<0.5 among the 396 non-jump nohook WETH climbers; if >20% they are a honeypot pocket and not a gap; if ~0% the 14 missed runners are real and a WETH rule is worth building.
- analyst2: speed-bucket filter at the 1M entry (P3 34% vs 15%).
Nothing ready for testmgr.