# Analyst1 idle run, 10 Oct 2026 (iter 27): where the runners really are now
Scripts s1-s7.py in /opt/crew/work/analyst1. Data: e1x-runners-v2 (27k pools, 21 Sep-9 Oct), live radar.jsonl (7-10 Oct), fm_mirror mc5.db (44 live trades + 1,100 skips), gold tapes, p4.pkl (NHW 388).

## Headlines
1. **The unwatched pool types are drying up or phantom.** Genuine runners (>=1M, ok flag, wash<.5, >=10 independent buyers), per day since 3 Oct: v3 8.9 (2.7 reach 5M), USDG nohook 2.1 (1.4 reach 5M, dying), hooked 0.3, native nohook ~0. WETH nohook 300k touches fell from 17-38/day (Sep) to 1-6/day; USDG nohook 24/day to 0-5. Native nohook "runners" (25 in 19 d) have 4-12 buyers, thousands of swaps, peak 1.0-2.1M and none above: script ladders, not a runner tier.
2. **Live 5-Min Cap: a 15-loss streak is normal, not a regime.** 44 live trades: 3 winners (29x, 19x, 13x), mean 1.77, win rate 6.8%. Bootstrap of the real trades over 100 trades: longest losing streak median 32, P(>=15)=99%. Median max drawdown 20 tickets, p90 32. Kelly fraction ~6% of bankroll per trade, ~1.5% without the best winner. Do not pause on streaks; size a ticket at <=2% of the bankroll.
3. **Ratchet/trailing exits on the live 44 trades hurt.** Replaying the real cap paths: stop .5 only 1.86; ratchet after 3x (trail .4-.7 or breakeven) 0.94-1.43. The three winners dip hard after 3x, so any tight trail cuts them. 10 of 44 touched 3x yet only 3 paid. Confirms #83/#4: wide stop is the point.
4. **NHW-S6 evidence is front-loaded**: 39 of 45 S6 trades are 22 Sep-2 Oct; 5-9 Oct n=6 mean ~1.5. Paper test nhws6 has 0 trades because NHW touches are now ~1/day. Not a bug, supply is gone.
5. **Runner-check overstates misses**: 7 of 11 non-hooked v4 "runners" in the last report fire all thresholds ($1M-$20M) at the same timestamp (first-print caps: Mouse, PERSUASION, SPY...). Ask research to tag them.

## The 10 iterations
| # | lens | test | verdict |
|-|-|-|-|
| 1 | exits on live data (ratchet stop) | real cap paths of 44 live mc5 trades | wide stop best, DISCARD |
| 2 | survival/hazard after 300k touch (stall exit) | share reaching 1M still, by minutes since touch, per type | flat hazard, no time-stop signal, DISCARD |
| 3 | deployer wallet nonce (launch hygiene) | v3 5.2k pools: real rate nonce<=5 4.7% vs >5 0.9%; given 300k touch age<=10 min 52% vs 31% (n=36) | MATURE-thin, data request ppdata |
| 4 | regime inside S6 (time decay, hour of day) | NHW S6 n=45 by day and hour | flat by hour, supply gone, INFO |
| 5 | venue lifecycle (new lens): supply per day by type | 300k touches and real runners per day | v4 nohook dying; v3 steady; INFO, redirect |
| 6 | live misses: cap_out skips | 891 matched checks: real rate by cap at +5 min | <250k: 5 real in 838; 250k-1M 37% (15 of 41); >1M covered by V3-1M-RIDE; DISCARD |
| 7 | wake-up burst after quiet (v3 fee100) | 588 bursts, enter next print +5s, tp3 sl.5 | 1.25 pre-cost, median .5, halves 1.18/1.32, DISCARD |
| 8 | risk of ruin/Kelly (new lens) | bootstrap live 44 | see headline 2, INFO to manager |
| 9 | inversion: phantom runners | runner-check thresholds same timestamp | see headline 5; task to research |
| 10 | tier census by type with independent buyers (new lens) | tiers p3/p5/p10 per type | v3 .44/.33/.20, USDG nh .53/.40/.21, hooks 27 only, native 1 |

## Data request (ppdata): deployer nonce + creator age for every pool
deployer.db (nonce at creation) only covers v3 pools since 6 Oct. Need: all pools in /opt/ppcopy/e1x-runners-v2.csv (27k, 21 Sep-9 Oct, all venues): creator address, tx nonce at the create tx, creator wallet first-ever tx block and first-funding source (CEX/bridge/EOA). Format: sqlite or csv in /opt/ppcopy/creator_age.csv with pool, creator, nonce, first_tx_b, first_funder. Test: P(real | 300k touch) by nonce<=5 per type and by week; pass = lift >=1.3x in both halves with n>=60 per type, otherwise kill.