# h3n plateau explanation (#705), auditor2, 10 Oct
Verdict: plateau mechanics accepted (haircut x imp at fixed R=0.10), h3n NOT verified.
- Explained: net 5.0x = gross ~10.3x x 0.94 x 0.51. Not a price-level artifact of our code.
- Still suspicious: gross ~10x at +595s in many different pools with identical cap_buy and 4 pre-swaps = same-template launches, i.e. one operator. A scripted pump-to-10x on a template is exactly what a rug-bot does; outside sellers near the exit are unproven.
- Open checks: (1) sellability at the exit (sell eth_call or Ben $5 test), (2) list distinct tx.from / funders across the 80 pools; if <=3 operators the "80 trades" is ~3 independent samples, (3) hand rebuild of 3 fills from chain data, (4) replay vs live on overlapping coins.
- Send me these when ready; will re-audit within 2h.

## Update #711 (auditor2)
Operator count received (testmgr2): 168 distinct tx.from on 168 trades (reserve>=0.1 subset), 227/233 overall. Single-operator risk cleared at tx.from level. Funder clustering not checked (cfund.db possible). Still open for VERIFIED: sellability at exit (Ben decision on sim vs real test), hand rebuild of 3 fills, replay vs live overlap. h3n stays not verified.