# KITRAMP (#651): early entry on the deterministic 32k-kit ramp, replay on the real swap tape
Data: /opt/sw/ppdata/r6nat5_{0..3}.db (ppdata #2970, PARTIAL: ~880 of 1100 kit pools, merge still running). Scripts: /opt/crew/work/analyst2/k5.py, k5b.py.
Price = 1/px^2 (pool sqrt price), relative to the first print of the pool.

## Facts
- 795 pools with >8 swaps: peak/first median 7.2x, p10 1.8x. Median 28 s to reach 1.5x (p90 38 s), median 108 s to 3x.
- Net ETH in the pool at the 3x crossing: median 2.30 ETH (p10 2.26, p90 2.38): constant = deterministic operator ramp. $10 (~0.003 ETH) is 0.1% of that depth: no impact.
- Price at tape end is 0.97x of the peak (median), p10 0.90: the pool holds near the top after the ramp, no visible collapse.

## Rule (replay, ent/exit each delayed by d seconds to the next print, 3% slippage each side)
Enter when price >= 1.5x first print (cap ~48k), buy size $10. Exit at 3x entry or 0.7x stop.
| delay d | n | mean | median | win | p10 |
| 3s  | 710 | 2.61 | 2.89 | 98% | 1.31 |
| 8s  | 707 | 2.58 | 2.85 | 97% | 1.24 |
| 15s | 688 | 2.59 | 2.84 | 96% | 1.22 |
tp2: mean 1.87 (97% win). Thirds by birth time (3 days): tp3 mean 2.47 / 2.63 / 2.61, win 94 / 98 / 99%. Stable.

## Caveats (why I do NOT call it verified)
1. A swap tape cannot show an LP removal. If the operator pulls liquidity while we hold, we get 0, and the replay would count the last print. ppdata LP-remove request (#630 / #415) is still open. The tape-end-to-peak ratio 0.97 says pools are not collapsing visibly, but a pull leaves no swap.
2. Detection: entry is ~28 s after birth. mc5 does not watch v4 native pools, so a live watcher on the PoolManager Initialize events for this kit (first print 32-34k, fee100, nohook native) must exist. ppfeeds feed exists? (kit pools: ~370/day.)
3. Earlier finding (#650): these pools never reach 1M. This is a 2-3x scalp, not a runner catcher. Money = trades/day x 1.5 x size; 700 triggers in 3 days = ~230/day.
4. Exit side outside-wallet check: sells have to go through our own router into a pool where the operator is the only other trader; sell depth is real (2.3 ETH) as long as no LP pull.
Verdict: READY FOR PAPER (shadow with real on-chain sells quoted via the own quoter), not live. Kill criterion: any paper trade where the pool is not sellable at the exit time, or mean < 1.3 at n >= 30.