# Audit 11 Oct 2026 (auditor2): flow60, wlink, breadth
Data: /root/tm/<id>/trades.csv (read 11 Oct ~08:55 UTC). Not done this round: hand rebuild of 3 fills, eth_call sell sim (no verdict needs it for the KILLs).

## flow60 (owner testmgr2) -> KILL (recommend)
- Forward (entry >= 10 Oct 02:00 UTC): n12, tp5 mean 0.48x (se 0.15), median 0.27, 2 wins. Whole file 59 pools: tp5 0.91x, median 0.01x (more than half the pools go to ~zero = pull/rug after our entry).
- Kill rule (n>=20, mean<1, upper80<1.05) not yet formal at n12, but upper 80% bound is 0.48+1.28*0.15 = 0.67 <1.05, replay itself 1.01x => no edge even in-sample. Paper "tp10/ladder" arms are the same trades.
- Not tested by the builder: outsider sellability, depth at our size. Not worth a slot: needs 0.005 ETH/day.
## breadth (owner testmgr2) -> KILL (recommend)
- Arm A all rows n45 mean 0.99x (se 0.21), median 0.44, 6 wins; forward A n3 0.57x. Arm B n28 1.18x median 0.45; mean carried by 4.65x tp hits (one pool repeated across arms). Kill rule "A < 1.1 over 30 closed" is met (n45, 0.99).
- Median 0.44x means a stop out is the typical outcome; no look-ahead issue needed to explain it. Replay 0.99x agrees: rule is weak.
## wlink (owner testmgr2) -> FIX (not verified yet)
- Forward since 10 Oct 11:50 UTC: arm A (cap<150k) n29 mean 2.36x, median 1.53, 15 wins; without the top 3 wins 2.10x. All caps n33 2.13x. 24 of the 29 are pools under 60k cap, mean 2.53x. Matches replay 2.25x, so no replay-vs-live gap. Exits: 12 tp, 11 time, 10 stop on all-caps forward; runners flagged 0 in forward (it wins by 5x scalps, not runners).
- Look-ahead of winner list: rule uses earlier-pool runner status only >=6 h after that pool started; a pool reaching 1M after 6 h would be mislabelled as a "winner" only if known later: small leak, ask builder to use runner status as of signal time.
- OPEN (blocks VERIFIED): (1) pull not modeled and 20-60k pools: our size vs real depth + outside sellers near tp5 exit not checked; tp5 on a 25k cap is only ~$125k, easy to be operator wash; (2) sellability (eth_call sim / hook) for these fee-100 pools; (3) winner wallets 6925 - are they one operator (same funder)? (4) n40 forward.
- Needed from testmgr2: add per-trade pool liquidity at entry and exit, a pull flag (reserve drop > 50% before exit), and outside-seller count in the 60 s before the tp exit. Re-audit within 2 h of resubmission; auto at n40.