# Audit v3r300 (auditor2, 13 Oct 2026, #2329)
Verdict: NOT VERIFIED, keep running, no FIX needed from testmgr2 yet. Re-audit at forward n>=40.
Note: the registry blocker "0 entries yet" is stale: 20 closed forward trades (20 distinct pools, src=fwd, trades.csv 13 Oct 11:10 UTC).

Forward (tp10, A1 main arm): n20 mean 1.67x, 80% band 0.65-2.69, median 0.00. 16 of 20 below 1x; 4 winners (3 tp at 9.7x, 1 time exit 3.9x); 5 pulls = 0 (25%).
tp5: mean 1.65x, 80% band 0.99-2.32, 7 hits. tp10sl3: 0.95x, stop arm loses.
Kill rule (n>=20, mean<1.0 and upper bound<1.05): not triggered. Fast-track (n>=40, mean>=1.3, lower bound>=1.0): not met (n20, lower 0.65).
Replay vs forward: replay n=27 is in-sample, pre-fee, claims 85% reach 1M and mean 2.63x. Forward tp10 hit rate is 15% (3/20) and about 20-35% for 5x. Replay hit rate sits far outside forward; the 85% is not reproduced. Mean is similar only because pulls and tp wins are lumpy.
Open checks: (1) sellability by eth_call on v3 fee-100 WETH pools (the UR sim reverts empty on WETH-quote, so use census senders + real payout comparison); (2) outside sellers near exit (needs sender census of tp exits); (3) pull share 25% with pull = 0 is the only loss modelled, tp fills at the exact tp price with no gap (optimistic); (4) hand rebuild of 3 fills; (5) replay audit with no look-ahead (reserve>=14 at +300s uses data up to entry, ok, but the 1.5%/side cost is assumed, not measured).
Money: ~4 entries/day expected, observed 20 in ~2 days = ~10/day over fast 20 trades; at $10 and mean 1.67x that would be ~0.01 ETH/day only if the mean holds, which the band does not support yet.
Next: wait n>=40 (about 2 days); then check mean>=1.3 and lower bound>=1.0, then run sellability/outside-seller checks.