# Analyst 1, iterate run 23 (10 Oct 2026). Scripts n1-n15.py in /opt/crew/work/analyst1

Focus: pools with $1M+ runners that we do not test. Clean runner = vol>=40k, swaps>=50, cap_flag ok, one label per token.

## Clean runner count (e1x, 21 Sep-8 Oct, tokens that touched $1M): 500
v3 WETH (mc5 universe) 196 | v4 hooked WETH 121 | v4 no-hook USDG 77 | v4 no-hook WETH 36 (mostly H3000 a80 decoys) | v4 no-hook native 24 | v4 hooked USDG 23.
P(reach 5M | touched 1M, not a jump): v3 WETH 0.28, v4 hooked WETH 0.21, v4 no-hook USDG 0.10.

## Result: V3-1M-RIDE (READY -> testmgr)
Buy a v3 fee-100 WETH pool the moment it first touches $1M cap (price*supply*ETHUSD), if the pool is <120 min old and first touched $300k at least 180 s earlier (not a jump). Market buy 3 s after the touch. Exit tp 5x, hard stop 0.3x, max hold 6 h; pull = total loss.
Sim (sim2 incl. price impact, fee, pull = zero, exit-depth check), 17 days 21 Sep-7 Oct:
- n=114 (6.7/day), mean 1.64x, median 0.30 (30% win), halves 1.68/1.61, without top 5: 1.49.
- pos 0.02 / 0.1 / 0.3 ETH: 1.64 / 1.60 / 1.52. tp8 sl.3 1.74; tp10 sl.25 1.77.
- positive on 12 of 17 days. fee-100 pools only: n=109 mean 1.72 (fee 1% pools n=5 mean 0.03).
- Age matters: <30 min 1.88 | 30-120 min 1.46 | >=120 min 0.58 (n=21, so excluded).
- No other filter works (gap 300k->1M, buy share, swaps/5min, unique senders: all non-monotone). Not added on purpose.
- Disjoint from live 5-Min Cap (that enters at +5 min under $1M). This is the "late runner" hunt: entry $1M, target $5M+.
Risks: 6h hold / 30% win rate = lumpy; sim uses the swap tape, last data 7 Oct, no OOS beyond that; 39% of these pools are pulled within 6h (exit must beat the pull; the sim already counts a pull before exit as zero).

## Discards
- 5-Min Cap on v4 no-hook USDG pools (cap 250k-1M at +5 min, big swap >=$20): n=84, tp3 sl.5 0.79, tp2 0.93, no fee. DISCARD.
- Buy at first $1M on USDG no-hook: only 29 smooth touches, P(5M|1M)=0.10. Not tested further.
- Wallet reputation (early buyers repeating across runners), USDG/SUSD all 13.5k pools, 5-fold walk-forward: <=9 test pools ever had a "good" wallet, 1 runner. In-sample on the 69 real touchers looked like 0.75 vs 0.36 but it was fitted on the same pools. Insiders rotate wallets. DISCARD.
- Sibling pools: 39% of runner tokens first run in a later pool of the token, but "hot sibling" lift (2.6% vs 0.5% on USDG no-hook) is the same token counted per pool. DISCARD (token-level label needed).
- Holder breadth at the 1M touch (runtrack rh_flow): coverage 29 of 114. Inconclusive, no request.

## Mature, thin
- USDG no-hook enter at first $300k touch (real pools that start <300k, n=60): tp5 sl.5 1.78, tp10 2.25, trail .6 after 2x 2.93 (halves 1.27/4.53). Too few; forward tape needed.