# KITRAMP (#651) rerun with the full LP-remove list (ppdata #2977/#2986), 18 Oct
Script k6c.py (= k6b.py on kit_lp_remove.csv, 981 of 985 swap-tape pools have a pull time). Entry at price >=1.5x first print (~29 s after first LP add), tp3 or tp2, stop 0.7x, delay d each side, 3% slip each side. Any fill (entry or exit) at or after the LP-remove time books 0x.
| d | tp | n | mean (pull modelled) | zero (pulled while holding) | win | mean without pull |
|---|---|---|---|---|---|---|
| 3s | 3 | 917 | 2.31 | 17% | 82% | 2.60 |
| 8s | 3 | 909 | 2.21 | 22% | 77% | 2.58 |
| 15s | 3 | 893 | 2.14 | 26% | 73% | 2.59 |
| 8s | 2 | 905 | 1.69 | 14% | 85% | 1.87 |
Reading: the pull costs ~0.4x of the mean but the edge stays: faster exit (small d) is worth more than the tp level. Pulls happen median 437 s after add, 15% inside 30 s, 30% inside 2 min; our hold is ~80 s.
Caveats for paper/shadow: (1) it is a rug farm: pool LP owner = deployer, so the pre-buy rug gate blocks it by design; paper/shadow only, any live use is Ben's call with that stated. (2) Sells in the replay are tape prints, not quoted sells: the shadow must quote real sells with the own quoter. (3) Not a runner catcher (peak never 1M). (4) Detection needs a PoolManager Initialize watcher on first print 32-34k native nohook fee100 (~370 pools/day; ~300 triggers/day).
Kill: mean < 1.3 at n>=30 or any non-sellable exit.