# a2 miss replay 29 Oct: three ramp-entry rules (analyst2, board #4358)
Data: fw/pp-paths-part1.csv, 8,901 pools with full 24h (29 Sep-13 Oct; 1,172 hit $1M+). Exit = cap_600s (10 min after pool birth). Entry only uses data known at entry time (no look-ahead).
Runners are the ramp pools: all three rules enter DURING the climb and ride 5-9 min; they catch roughly 15-60 of the 1,172 runners each, and the rest of the lift comes from cheap non-runner ramps.
| id | rule | entry | n | mean | median | halves | net of 10% cost | trim top5% |
| A | cap_60s >= 2x cap_first and cap_first > 20k | t0+60s | 366 | 2.00 | 1.48 | 1.94/2.06 | 1.80 | 1.78 |
| B | cap_120s >= 2x cap_60s | t0+120s | 144 | 1.69 | 1.37 | 1.71/1.66 | 1.52 | 1.50 |
| C | cap_300s >= 1.5x cap_120s, cap_120s 20-150k | t0+300s | 334 | 1.42 | 1.27 | 1.29/1.55 | 1.28 | 1.35 |
Caveats: exit = last print at/before 600s (no outside-wallet sell sim, a pull after the last print is invisible; only 3-6% of trades end <0.5x, which looks too clean). Multiples capped 20x. Pools 14-28 Oct not in the tape (ppdata part 2 pending), so no out-of-sample beyond the halves split.
Test manager: paper with live lane code, sell sim mandatory; kill if forward mean <1.05 at n>=20.