# Analyst 3 idle run 4 (10 Oct 2026): where the unwatched runners really are

Scripts: work/analyst3 n1-n9.py. Data: radar.jsonl (7-10 Oct crossings), r6weth + v3hist/v3live (v3 tape), e1x + deployer_raw.db (v4 pool-init tx_from/tx_to), btceth_1m.csv.

## Headline
Of 53 v4 no-hook 1M crossings on native/USDG/WETH in 3.5 days, 51 arrive as a single print (first seen cap >= 0.7M). Only 2 are climbs, with $0-1 depth. Over 9 days (e1x + deployer_raw) 226 real v4 climbers: 109 are the direct-PoolManager WETH hooked kit (H3000, already in paper h3k), 35+20 are launcher 0x58da no-hook fee-100 WETH/native (N300/NHW family, plateau at 1.0-2.1M), 33 USDG no-hook (0x58da), 9 other. The unwatched v4 space holds no hidden big-runner pocket.

## 10 iterations
1. Venue census, radar 3.5 d (focus: v4 nohook). 51/53 single-print; 17 non-jump 300k touches on native: 0 reach 1M; USDG 1 of 4. v3: 37 climbers to 1M. DISCARD (confirms #57, #96).
2. Order-flow toxicity, router fingerprint, metronome (new lens), v3 mc5 band n=330: inter-swap CV, distinct routers, direct callers, buy-size growth flat. First buy delay, buy/sell imbalance and swaps in first 5 min are one factor (rank corr -0.76): swaps<10 in 5 min 103 pools tp5 0.96 hit5 1% vs 227 pools 1.37 (halves .99/.92 vs 1.44/1.30; weeks 39/40/41: 1.02/.90/.91 vs 1.57/1.20/1.36). Confirms analyst2 #91, wider than the dust guard. MATURE (mc5 gate: require >=10 swaps by +5 min, +0.13x).
3. Regime, BTC/ETH minute data (btceth_1m.csv), 1/4/12/24 h returns and drawdown from 24 h high, terciles on 330 band pools: all 1.1-1.4, no order, halves flip. Also "afterglow" (1M/5M runners in the previous 1/3/12 h): noisy. DISCARD. 15 losses in a row is routine per ruin math (#84).
4. Mint-to-PoolManager factory 0x8366 (tx_to / creators.factory), no-hook WETH fee 3000 (focus): 32 pools "11 reach 1M" but 0 real; they are the decoy twins of the H3000 hooked pools (#62). The real runners sit in the hooked sibling (h3k). DISCARD.
5. Seeded 1-2.5M pools on v4 (focus): 6 pools born at 1.0-2.5M cap, 0 reach 3M; 6 born >=2.5M are PERSUASION/CASHCAT/SPY-type phantoms. DISCARD.
6. Seller side, early wallets (first 30 s buyers) net out by +5 min: flat (tp5 1.29/1.26/1.34 by terciles). DISCARD.
7. Where v3 climbers to 1M fall (radar 7-10 Oct): 13 reach 1M < 5 min (4 reach 5M), 17 in the mc5 band (5 reach 5M, all 3 of the 20M), 7 late bloomers after 5 min (0 reach 5M). mc5 band holds 46% of climbers and all 20M ones. INFO.
8. Seed-cap preset as launcher fingerprint (v3 fee100, v3hist+v3live 5.6k pools, 12 Sep-10 Oct): "hot kit" (same 8% cap0 bucket produced >=2 real 1M pools 24-96 h earlier, bucket <40 pools) gives P(1M) 14-25% vs 1-3%, both halves. But entry at +5/15/66 s returns are worse (hot 1.66/1.70/1.53 vs cold 2.23/2.39/2.42; tiny-cap tp5 method is optimistic in absolute terms, comparison within method). The lift is the fast jumpers that are gone before entry. DISCARD as entry filter; kit persistence is real (days), so deployer data (#509) could still make a pre-entry flag.
9. Recent launcher/deployer census v4 (deployer_raw tx_to/from x e1x): no new launcher; top tx_from (8 climbers each) are fleets inside known buckets. INFO.
10. DATA REQUEST to ppdata: swap+liq tape for the 56 real no-hook fee-100 WETH/native 0x58da climbers (peak 1.0-2.1M, 5.6/day average 22 Sep-9 Oct) and 366 touched-300k-no-1M controls: none of them is in r6tape.db. Question: do these plateaus get pulled, how deep is the exit, what does an N300 with tp2.5 net? Pass: mean >= 1.3x after 3% slip with pulls <15%; kill: pulls >40%. Prior is low (13% of 300k touchers reach 1M, peaks barely 3-7x).

Nothing ready for testmgr. Candidate for analyst2: mc5 gate swaps>=10 by +5 min (iteration 2).