# analyst1 idle run 48 (10 Oct 2026): nohook v4 weth/native, late-sell arm (#435) stress tests
Hunt: entry +315 s, entry cap <150k, target 1M+ runners (via tp3 on the tape). Scripts: work/analyst1/q1/h1-h7.py (sim = pull-aware r6 tape sim from p1/g3.py).

| # | Lens | Result | Verdict |
|---|---|---|---|
| 1 | Exits (trailing, ladder, tp2/5/10, stop .3/.7) | flat tp3 stop .5 best: 1.194 n748 (halves 1.09/1.30); ladders/trails 0.90-1.11 | discard |
| 2 | Regime: births in prior 60 min (canon) | quartiles 1.38/1.25/1.13/1.00 monotone | looked good |
| 3 | Regime replication on births.db | non-monotone, skipped half1 beats kept | discard (artefact of canon) |
| 4 | VPIN analogue: USD-weighted sell share | non-monotone, adds nothing to count share | discard |
| 5 | Early/mid sell share, trend | noisy, halves disagree | discard |
| 6 | Capacity, lag, cost | pos .005-.2 -> 1.20-1.14; lag 15/30/60/120 s 1.19/1.20/1.14/1.11; cost .04-.12 -> 1.19-1.09 | supports #1294 |
| 7 | Seller identity via Transfer logs (237 pools) | v4 sell payer = router, no wallet identity | discard |

Side result: late-sell filter replicates on a fresh random 237-pool subsample (kept 1.32 n161 vs skipped 0.85 n76).
Nothing new for testmgr; #435 (testmgr #1294) gets the capacity/cost evidence. Open data blockers unchanged: origin EOA for v4 swaps (#881), nohook liq tx.from (#1247).