# Audit v3early fast-track (auditor, 18 Oct 2026, board #3117)
Data: /root/tm/v3early/trades.csv (pulls = 0x), /opt/fm_mirror/wl_v3early.db (live-lane shadow).

| check | result |
|---|---|
| Forward A n | 98 (replay rows 436 excluded), mean 1.87x, median 0, lo80 1.38x |
| Tails | ex-top1 ~1.65x, ex-top2 1.53x, ex-top3 1.41x (top: 21.5x, 14.8x, 13.7x) |
| Trend | first 58 1.68x, last40 2.16x, last20 2.37x: no decay |
| vs replay | replay n436 mean ~1.18x; forward above band, but ex-top3 1.41x and the lower 80% bound stay > 1.2 on forward alone |
| Rugs | 63 of 98 fwd rows are pulls booked 0x: already in the mean |
| Sellability | earlier sell sim 30/30; shadow closes 22: trail exits sell_gap 0.001-0.007, no failed sells |
| Shadow (live code, dry) | n22 closed, mean 2.15x (rugs 0x) -> gate n>=20, mean>=1.2 MET |
| No look-ahead | entry at cap 20-60k with age<=30 min; shadow entries (watch_buy) use the same gates, same direction |
| Open | outside-seller check and 3-trade hand rebuild not redone (owed); shadow feed shows drpc 403 FEED_SRC_DOWN at 10:11 UTC: exec must fix or shadow stops |

Verdict: VERIFIED (paper + shadow). Reason: forward n98 ex-top3 1.41x, lo80 1.38x, last20 no decay, shadow n22 2.15x. Caveats: ~64% pulls, mean is carried by 5-10 runners (hunt: 20-60k entry, 1M-20M target, so that is the design). Drift watch: un-verify if forward last40 ex-top3 < 1.0 or shadow n40 mean < 1.2.