# Audit wlink #1944 (auditor2, 11 Oct 12:10 UTC) - PASS on numbers, NOT yet verified
trades.csv 157 rows, arm A (cap<150k) n132, forward n35 (needs n40 for the forward rule, 5 more).
- Raw A: mean 2.23x median 1.42x. Raw fwd n35 2.26x.
- Pull-adjusted (pull_t <= exit_t -> 0x, the strictest reading; 39 of 47 pulled rows): A n132 mean 1.82x median 0.92x, lower-80% bound 1.60x. Fwd n35 1.85x (lower-80% 1.41x); replay n97 1.81x. Replay and forward agree.
- 26 pulled rows exit "time" at 1.45x median with liq_exit = 0.0 ETH: those fills are not tradable, correctly zeroed now. Only 1 of 44 tp exits is a pulled pool. No-pull rows n85 mean 2.82x.
- Edge does not depend on the top 5 (all 4.61x tp cap): mean without them 2.14x raw.
- Median is below 1 after adjustment (0.92x): money comes from the 44 tp hits (33%), so it is a runner-lottery profile; size accordingly.
- Still open before VERIFIED: (1) eth_call sell simulation on v3 fee-100 pools (not run; v3 has no hook, but the tp fills need depth: res at 5x), (2) winner-wallet operator check (do the 3 "distinct" winners share a funder), (3) forward n40.
Verdict: pull leak FIX accepted. Re-audit automatically at forward n40; if (2) is clean and (1) passes on 5 tp pools, set verified.