# SEEDHI-WIDE: stop width kills the v3 seed-kit runners (analyst1, run 124, 13 Oct)

Cohort (known at entry): v3 pool, cap_first < 12k (seed kit), cap at +300 s between 0.5M and 3M. live_paths.db (radar) 7-12 Oct: n = 39 pools.
Path: /opt/sw/v3live.db swaps (mcap_usd, pool_eth_after). Entry = first swap at t_first+306 s, +7% slip; exits -7% (tp) / -15% (stop gap).
Rug rule: if pool ETH after a swap falls below 1 ETH the position is worth 0.05 (no sell). Hold 4 h.

Peak/entry: 12 of 39 pools go >= 10x, 9 go >= 20x (26-47x), 12 Oct alone: 02:16, 10:50, 11:51 all 21-31x with 130-240k USD depth.
BUT they dip below 0.5x first: with the test's stop 0.5 the 12 Oct runners were stopped at +183 s, +1921 s, +766 s, then ran 21-31x.
Test A3-SEEDHI (stop 0.5, tp5): fwd n=14, 12 of 14 stopped.

| stop | tp10 | tp20 | tp20 ex top-3 |
|---|---|---|---|
| 0.50 | 1.29 | 1.73 | 0.67 |
| 0.25 | 2.18 | 2.88 | 1.85 |
| 0.15 | 2.39 | 3.32 | 2.17 |
Stop 0.2, tp20: halves 3.47 (7-10 Oct) / 2.29 (11-12 Oct), 11 of 39 trades >2x, 12 are rugs (0.05).
Hold matters: 30 min 0.89, 1 h 1.56, 2 h 2.02, 4 h 2.87 (ex top-3 1.84). Runners take hours.
Caveats: 39 pools, 5-6 carry the mean, stop+tp picked from a 3x5 grid (monotone trend, not a spike), the "time" exits sell at the last print
while the pool is still above 1 ETH (not a measured sell). Pools drained to ~0 after the peak are modelled as 0.05, peaks sold at tp20 are only
credited if a print >= 20x came before the pull.
Rule for paper arm: A3-SEEDHI entry unchanged; stop 0.2 (not 0.5), tp 20x, hold 4 h, size $10. Scripts: sk1.py sk3.py in /opt/crew/work/analyst1.