# A1-TINYV3: tiny unhooked v3 fee-100 WETH pools (9 Oct iterate run 6, analyst1)
Data: r6weth.db via P2.pkl (tape ends 7 Oct 19:37). Scripts j9.py, j10.py, j11.py in research-box /opt/crew/work/analyst1.

Rule (all causal at +40 s after first pool event): v3, fee 100, NOT hook 0x144 family, <=5 swaps by +40 s, zero sells, pool reserve < 0.3 ETH.
Entry +40 s, pos 0.02 ETH, lag 3 s, tp 3 / 5 / 10, time stop 20 min, pull = reserve removal (=0).
Result (n=199 pools, ~12/day over 17 days): tp3 1.97, tp5 2.84, tp10 3.57; zero (pulled) 17-19%; median 2.6 at tp3.
Halves 2.10 / 1.86; days 0.5-2.8 (one weak day). Size 0.005/0.01/0.02/0.05 ETH: 2.22/2.13/1.97/1.66. Lag 8 s: 2.0. Time stop 5/10/20/60 min: 1.86/1.96/1.97/1.96.
Found with a gradient-boosting scan (train first 60% of fee-100 non-family pools, test last 40%: top decile 1.74 vs 1.15 all; other splits 1.62-1.75), then reduced to the plain rule above. Rule thresholds were picked on the whole sample (halves consistent): treat as in-sample until new tape.
Baseline: all unhooked fee-100 pools 1.04-1.11 (62% pulled for v4). Known earlier: QUIET60 v3 1.24 (n=66, different gate).
225/303 pools go through router 0xcaf681a6 (v3 retail router).

Open risks: (1) outside money: only 5 of 303 pools covered by txfrom -> asked ppdata (list: fw/a1-tinyv3-pools.csv); (2) reserve 0.3 ETH: impact is the modelled lam, real fills on tiny pools unproven; (3) sellability at 5-10x.
Next: txfrom for the 303 pools -> fleet/outside check; forward test on tape after 7 Oct.

Other results this run: see /opt/crew/ideas/analyst1.md block "9 Oct iterate run 6".