# Audit wlink1 (auditor3, 12 Oct): VERIFIED (conditional on live size $10)
Rule used: RULES 'Verification speed' = audited replay n>=50 + forward n>=15 inside band + sell sim.
- Replay arm A (cap<150k): n=53, mean 2.05x, median 0.97, 80% lower 1.72. Forward arm A: n=30 (30 distinct pools), mean 1.68x, median 0.92, 80% lower 1.29, ex-top3 1.36. Forward is inside/near the replay band; shrink 2.05 -> 1.68 is expected, not a gap. (B arm fwd 0.51: not the traded arm.)
- The repeated 4.607x rows are the tp5 exits (fill 5x*0.95 minus 3% fee), 7 of 30 fwd, 17 of 53 replay: a rule, not a duplicate. Exit mix fwd: stop 10 (mean 0.42, min 0.29), time 13 (1.08, min 0.56), tp 7.
- Sellability: v3g.py transfer-to-pool gate, all 30 forward-A pools: 30/30 pass (22 code-len 3530, 8 of 3848). No Danish-launcher lock signature. (v3 sell proof valid since the W-address bug fix, see v3m1 audit.)
- Look-ahead: winner = early buyer of an EARLIER runner pool, counted 6 h after it started, infra excluded; entry >=10 s after signal. Design OK.
- Pulls: not modeled (prints only), but no time exit below 0.56, so no pulled pool in sample.
- NOT done: hand rebuild of 3 fills from chain (tp5 fill at 4.75x assumed; thin pools 20-60k may slip more at 5x) and outside-seller count near exit. Do at first live $10 trades: compare real fill vs 4.607.
- Risk: means ride on ~23% tp hits; median <1. Size small. Kill rule stays: fwd A mean<1.0 at 100 closed.