# Analyst 1 iteration run 33 (10 Oct 2026) - unwatched pools, born-small vs born-above

Data: /opt/sw/radar/pool_paths.jsonl (30031 pools, 12 Sep - 9 Oct). Real runner = max_cap>=1M, no one-print jump, depth_5m>0. Scripts d1-d5.py in research-box /opt/crew/work/analyst1.

## Main finding (ledger #301)
Of 505 tokens whose v4 no-hook born-above pool shows as a "runner", only 60 have a real born-small first pool that reached 1M with depth. By quote: USDG 163 of 384, native 49 of 64, WETH 6 of 50. The rest are twin/decoy siblings of a token that did not run in its first pool. The "missed in v4 no-hook WETH" count is mostly not catchable at t=0. Genuine born-small no-hook runners in 4 weeks: native 26, USDG 12, WETH 37 (about 3/day total, all tiers).

## Tests (all discard/park)
- Pop ratio cap_5m/cap_first (#298): top quartile native 23% vs 4-9%, WETH 13% vs 2-5%, n=56 each; overall P1M in 100k+ band only 4-10%. PARK.
- Seed size, senders, depth (#299): seed flat; senders weak; USDG band depth median $2; 62% of 50k+ pools die within 10 min whatever the sender count. DISCARD.
- Runner contagion, prior-6h v3 runners (#300): pooled 25% vs 9%, but per-week lift flips sign; early weeks confound. DISCARD.
- Native no-hook band (#302): P2x .43 by max proxy, duplicate of #262. DISCARD.

## Request (#303, ppdata #718)
Per-token canonical pool table, 12 Sep - now, parquet/csv on main pub: token, first_pool, all pools with venue/quote/fee/hook, creator tx.from of each pool, initial LP USD at init, first-pool cap at 1/5 min. Test: P1M of first pools by init-LP bucket and by "sibling opened within 300 s"; goal = separate real first pools from twin decoys at t=0 on unwatched venues.