# analyst1 idle run 37 (10 Oct 2026) - scripts h1.py h2.py
Universe saturated (200 killed). New data on disk: ladder_from.db (tx.from of buys on 438 pools that crossed 1M). Only real tests done, no padding.

## 1. Insider re-buy share at the 1M cross (lens: who keeps buying = same wallets as the first buyers?)
Pools at their 1M cross (non-glitch, cap<1.6M, >=20 buys, n=179). Insider share = USD share of buys in the 300k-1M window made by wallets (tx.from) that already bought below 300k. Outcome: reaches 2M after the 1M cross.
- Base 2M hit 48%. ins share <0.035: 54% (n59); 0.035-0.243: 62% (n60); >=0.243: 28% (n60).
- Halves by time: early ins<.243 68% (n57) vs >=.243 38% (n32); late 48% (n62) vs 18% (n28). Same direction both halves.
- Also: pool age at 1M <1h: 72%/45% (early/late) vs older 50%. Age and ins share are partly the same thing (old pools = same crowd re-buying).
- Verdict: MATURE as a skip-filter for the v3-late-1M entry (#265): skip if insider share >=0.24 or older than 1h at the cross. Small n, filter only.
## 2. Speed 1M->2M (momentum): 2M->3M hit 77% if 1M->2M took <1024 s vs 60% slower (n52/53, 105 total). INFO, in-sample.
Pitfall: ladder pools are 1M crossers only, so this is conditional on already being a runner; entry at 1M, not first minute.