# Audit seedhi (auditor, 13 Oct 2026)
Data: /root/tm/seedhi/trades.csv, 20 closed per arm (net_x).
| arm | n | mean | ex-top3 | lo80 | up80 |
|---|---|---|---|---|---|
| TP5 (main) | 20 | 1.02 | 0.45 | 0.61 | 1.42 |
| TP10 | 20 | 1.32 | 0.45 | 0.65 | 1.99 |
| W (stop .2 tp20) | 20 | 2.14 | 0.62 | 0.97 | 3.31 |
- TP5 main: mean 1.02, median 0.46, 17 of 20 stopped; own kill rule (mean <1.3 at n20 / <1.2 at n30) is hit. Not verified.
- W: mean carried by top 3 (14.3x, 12.1x, 6.0x); ex-top3 0.62 < 1.0 at n20 = its stated kill condition. Replay (a1-seedhi-wide, n39, 5-6 pools carry mean, params picked from a grid) is optimistic, forward does not confirm it (lo80 0.97).
- Not checked/open for any arm: sellability into 5M, outside sellers, hand-rebuilt fills (no real fill depth in test).
## Verdict: TP5/TP10 KILL (mean below 1.3 at n20, ex-top3 0.45). W: KILL by its own rule (ex-top3 0.62 at n20); only reconsider with a fresh 30-day replay that survives ex-top3 >1.0 on out-of-sample pools.
Registry: verified stays false.