# A1 iterate run 19 (10 Oct 2026): 10 iterations, 1 data request, nothing ready for testmgr
Scripts: /opt/crew/work/analyst1 ca1-3 cb1-4 cc1-2 cd1-2 ce1-3 cf1-2 cg1 ch1 (research-box).
New lenses (not in either ledger): competitor-bot census, venue coverage audit (dark factories), attention boards (GeckoTerminal trending), crew labels (pretag), cross-pair arbitrage on twin v2 pairs, denomination (USD vs ETH bands), root reputation over the whole population.

1. BOT CENSUS (crowding). Early buyers (first 10 s) that were early buyers in >=3 earlier pools (v3live, n=344 enterable pools, tp5 30 s): repeat-bot count 0/1/2-3/4+ -> 1.43/1.42/1.72/1.40; halves 1.54 vs 1.45, 1.47 vs 1.37. Flat. DISCARD.
2. DARK VENUES. 10 minutes of chain-wide v2/v3 Swap logs (6,000 blocks): 72 v2 and 241 v3 emitting pools. Factories: all active new-pool factories (v2 8bceaa/0d1ebb, v3 1f7d75/e0c4ce/e51960) are in pools_since_68m. Three v3 factories not in the list (1ac9db, 16494a, aa5865) trade odd fee tiers (40-9000) but made 2 pools in 17 h. No missing venue. DISCARD.
3. AVERAGE TICKET in the mc5 cell (buy USD / buyers, 300 s): bottom quartile (<$3 per buyer) 0.88 vs 2.49 (n=84/226), both halves. But harness replay already removes them (1 of 195 replay trades below $3). Same as run 15. DISCARD.
4. LIQUIDITY / CAP in the mc5 cell: net reserve flow quartiles 0.92/1.66/2.58/3.85 but replay only has 8-13 of 183 below 2 ETH (pool_quote_check covers it). DISCARD.
5. BLOCK-0 SNIPER FLIPS (b0 buyers selling within 5 s): n=5 in cell, 2 in replay. Too rare. DISCARD.
6. ATTENTION BOARDS (trending.db, GeckoTerminal trending_5m, 1,547 tokens): median age at first appearance 50 min (only 8 of 180 matched pools within 10 min). Entry at first appearance, age 15-60 min: +30 min 1.47 n=46 vs same-age baseline 1.34 (baseline is survivor-biased). No edge, too late. DISCARD.
7. CREW LABELS (pretag.db, 6 crews): matched 143 of 1,855 native pools; per-crew means 0.43..3.67 with n=10-27, noise level. THIN, parked (needs longer history, see request).
8. TWIN-PAIR ARBITRAGE (47,331 tokens with 2 v2 WETH pairs in r6tape, exact CP maths, 0.3% fee): only 9 of 46,572 twin sets ever diverge >5%. Twins move together. No arb. DISCARD (closes the parked dual-pool idea for v2).
9. DENOMINATION (cap bands in ETH instead of USD): ETH range in the window only 2388-2782; 12 coins differ, no effect. Cell sweep: lower bound 400k gives 2.37 n=196 vs 2.05 n=310, known hump, fewer trades. DISCARD.
10. ROOT REPUTATION across the whole population (crew.db root of creator; P2 pools with root 1,757, 53 roots; expanding prior mean of earlier pools of the same root, >=5 prior): quartiles 0.54/1.06/1.28/2.37, but that is mostly family (h100/hook roots). Inside u100 (13 roots, n=1,179): terciles 0.55/1.02/1.18, halves H1 0.42/1.06/1.28, H2 0.68/0.98/1.08. One root (298 pools) at 0.43; others 0.8-1.27. Skip root prior<0.63: keep 1.10 (n=786) vs drop 0.55. Inside h100 and v4 fee 3000: flat. Stable but u100 stays ~1.1 at best: MATURE, not ready. REQUEST dataeng (below).

Data request (dataeng): fw/a1-iter19-dataeng.md.