# Audit ns812 (auditor, 30 Oct 2026) - NOT VERIFIED (FIX)
Numbers (/root/tm/ns812/trades.csv): replay n2124 mean 1.53 ex3 1.52, zero-outcomes 0; forward n22 mean 1.29 ex3 1.20, max 2.11, <1x 2.
- Forward vs replay: 1.29 vs 1.53, no look-ahead flag in the rule (buyers_60s/depth filters dropped). Forward is inside a plausible band but n22 is thin (needs 40 to bound it).
- Too-good flag: 0 losers below 0.1x in 2124 replay trades and "pull=0" is an assumption, not measured. Pools skipped "pulled before entry" (4) show pulls exist; pulls inside the 60 s hold are not modeled.
- Sellability: native nohook pools, so no hook code risk, but no eth_call sell sim and no outside-seller count at the exit second have been run. OPEN.
- Hand rebuild of 3 fills from chain: not done. OPEN.
Verdict: not verified. To verify: (1) outside-seller count near exit + sellsim on forward trades (testmgr/dataeng), (2) a pull-within-60s rate on replay (should be <5% or modeled as 0x), (3) forward n>=40 with ex3 >=1.1. Else retire at decide-by 3 Nov.
# s795 #4370 (auditor2, 29 Oct 07:25 UTC): VERIFIED (arm A, entry sell-sim gate mandatory)
- Forward A matured (>6h, on time, quiet counted) n35: mean 1.45x, ex-top3 1.32x, lower80 1.14x, tp 17/35. Replay A on-time n836 1.22x sits inside the forward 80% band (1.15-1.75). Earlier 2.16x/tp 82% was censoring; resolved.
- Sellability: entrysim 41 rows, 11 fresh (<=303 s): 10 PASS (exit ratio 0.99-1.02 at 0.1 ETH), 1 BLOCK. 30 rows STALE/late BLOCK = detector lag, not token result. Gate must block on BLOCK and on delay >303 s.
- Condition for live lane: exec's entry sim (p1/p01/e01) runs before every buy, stale = skip. Ben's $5 sell test only confirms.
- Watch: unverify if matured A last20 mean <1.0 or fresh BLOCK rate >30%.

## Re-audit 30 Oct ~15:30 UTC (auditor)
- Forward n40: mean 1.298 (bootstrap 80% 1.22-1.39), ex3 1.25, max 3.29, min 0.94, last20 1.30; replay 1.53 -> forward inside the replay band, criterion (3) MET.
- (2) Pull within 60s: measured 0/2146 (testmgr fix-ns812.md) -> MET. Zero forward trades under 0.94x in 40 is still unusually clean; keep watching.
- (1) Outside sellers: 69% of trades see a sell in the hold; sender is the router so distinct-sender is unreliable. No eth_call v4 sell sim from a fresh wallet. OPEN.
- Hand rebuild of 3 fills from chain: OPEN.
Verdict: NOT VERIFIED, one step away. Need: v4 swap sell-simulation (eth_call from a fresh non-deployer wallet via the v4 router/PoolManager) on ~5 forward pools at exit time, plus 3 hand-rebuilt fills. Else retire 3 Nov.

## Re-audit 30 Oct ~21:00 UTC (auditor, #4535)
- Forward n51: mean 1.273, ex3 1.20, <1x 8 (no longer suspiciously clean), pull 0. Inside replay band; criteria (2),(3) still MET.
- entrysim.csv: 9 rows, 4 PASS (exit ratio 0.997-1.013 at 0.1 ETH), 4 BLOCK, 1 INCONCLUSIVE. All checks ran 580-1700 s after entry (late, detector lag), so BLOCKs are not a token result; fresh (<=303 s) rows = 0. Not enough for sellability.
- Hand rebuild of 3 fills: still OPEN.
Verdict: NOT VERIFIED unchanged. Need entrysim on >=10 FRESH (<=303 s) forward pools with PASS rate >=70%, plus 3 hand fills. Else retire 3 Nov.

## Re-audit 30 Oct 17:20 UTC (auditor, #4559)
- Unchanged: fwd n~51 1.27 ex3 1.20; entrysim 25 rows, 0 fresh (<=303 s); hand fills still open. NOT VERIFIED; retire 3 Nov unless fresh entrysim n>=10 PASS>=70% + 3 hand fills arrive.

## Re-audit 31 Oct ~04:20 UTC (auditor, #4571)
- Unchanged: entrysim 35 rows, 0 fresh (<=303 s), hand fills open. NOT VERIFIED; retire 3 Nov unless fresh entrysim n>=10 PASS>=70% + 3 hand fills arrive.

## Re-audit 31 Oct ~05:00 UTC (auditor, #4572)
- Unchanged: entrysim still 35 rows, 0 fresh; hand fills open. NOT VERIFIED; retire 3 Nov unless fresh entrysim n>=10 PASS>=70% + 3 hand fills arrive.

## Re-audit 30 Oct ~21:00 UTC (auditor, #4535)
- Forward n51: mean 1.273, ex3 1.20, <1x 8 (no longer suspiciously clean), pull 0. Inside replay band; criteria (2),(3) still MET.
- entrysim.csv: 9 rows, 4 PASS (exit ratio 0.997-1.013 at 0.1 ETH), 4 BLOCK, 1 INCONCLUSIVE. All checks ran 580-1700 s after entry (late, detector lag), so BLOCKs are not a token result; fresh (<=303 s) rows = 0. Not enough for sellability.
- Hand rebuild of 3 fills: still OPEN.
Verdict: NOT VERIFIED unchanged. Need entrysim on >=10 FRESH (<=303 s) forward pools with PASS rate >=70%, plus 3 hand fills. Else retire 3 Nov.
- Update 31 Oct: fwd n20 mean 3.95x, ex3 2.05x, capped at 5x 2.00x, 12/20 zero (pull). Mean is positive but still no sell-depth sim at exit, no outside-seller count, no hand fills (0/3). Zero-trades are pulls modeled as 0x, so rug share is priced in; the open risk is whether winners' exits are sellable. NOT VERIFIED; needs testmgr items above. Decide-by 3 Nov.

## Re-audit 31 Oct ~05:45 UTC (auditor, #4586)
- Forward n2214 all-src mean 1.52 (ex3 1.51). entrysim 43 rows, 3 fresh (<=303 s), all 3 BLOCK (p1_ok 0, p01_ok 0) -> fresh PASS rate 0/3, below 70%; need n>=10. Hand fills still open.
- NOT VERIFIED unchanged; KILL-lean if fresh BLOCKs persist. Retire 3 Nov unless fresh n>=10 PASS>=70% + 3 hand fills.