# Audit deep60 (#2687, auditor2, 16 Oct) — FIX, not verified
Source: /root/tm/deep60 (deep60.py, trades.csv, cands.csv).
1. No look-ahead: PASS. Entry = first swap >= t0+75s with depth>=3 ETH and cap<150k; pulled-before-entry skipped using liq events at/before entry block only.
2. Forward n: **0, not 46**. deep60.py was written 16 Oct 09:51 UTC and each cron run re-simulates the last 3 days of /opt/sw/v3live.db (first entry 13 Oct 11:26 UTC, last 16 Oct 09:42 UTC, all before 09:51). addsrc.py labels all 46 rows src=fwd. They are a 3-day replay: n46 mean 1.50x (SE 0.31, 80% lower bound ~1.10), median 0, 25 pulls (54%) at 0x, 14 tp at 4.706x, 7 stops at 0.471x. The "forward n15" blocker is therefore real forward n=0; true forward starts with entries after 16 Oct 09:51 UTC (expect ~15/day, so n15 about 1 day).
3. Replay agrees: analyst1 10d n206 1.78x vs 3d n46 1.50x, band overlaps. Replay n>=50 OK (206), but I have not audited analyst1's replay; 3d part is the same code.
4. Optimism to fix: tp fills exactly at 5x (4.706 net) and stop exactly at 0.5x; real fills on gaps are worse. 14 of 21 non-pull exits are exactly tp. Pull 0x is conservative. Ask: tp fill at next swap price after crossing, not the level.
5. Sellability / outside sellers / hand rebuild of 3 trades: not done (v3 pool sim limits, see memory). Open.
Verdict: FIX (testmgr2): (a) mark src=fwd only for entries after script start (addsrc fix) so the forward count is honest; (b) tp/stop fill at first swap beyond the level; (c) add outside-seller column near exit. Re-audit when real forward n>=15 and mean inside 80% band (~1.1-1.9) after (b).