# Audit wlink fast-track (auditor, 11 Oct 19:xx UTC) - FIX, not verified
- trades.csv n=156 (arm A cap<150k n131). Forward A since 10 Oct 11:50 UTC: n34 mean 2.30x median 1.57x (needs n40 for the forward rule).
- Pull split (A): pull=0 n84 mean 2.85x (median 4.61x); pull=1 n47 mean 1.16x. 31 pulled trades exit "time" at 1.42x: if the pool was pulled, a 1.4x time exit is not credible unless the pull came after our exit. Need pull time vs exit time per row.
- Outside sellers: rows with <=2 outside sellers in 60s n22 mean 1.40x, so edge depends on exits being absorbed by outsiders.
- Not done: eth_call sell simulation (hook=0?), hand rebuild of 3 fills, replay vs forward, winner-wallet operator check.
- Mean 2.04x is real only if the no-pull population is the tradable one. Realistic estimate: between 1.2x (pull-adjusted) and 2.85x.
- FIX (testmgr2): (1) add pull_t and show pull before/after exit; if pull before exit, set net_x to 0 for pulled rows, (2) sellability sim per pool, (3) decide-by n40 forward. Re-audit at forward n40 (6 more trades).