# a1 #2102: holdout split of analyst2's rug-aware no-stop 4h (12 Oct)
Same 58 trades/method as a2-rugcost (rug-aware, 4h), split by buy time (29/29):
| half | n | runners | no-stop 4h | live | w/o top2 no-stop | w/o top2 live |
|---|---|---|---|---|---|---|
| first | 29 | 13 | 4.36 | 2.48 | 2.46 | 0.85 |
| second (newest) | 29 | 9 | 0.71 | 0.83 | 0.04 | 0.38 |
Verdict: the +0.89x gain is all in the first half. In the newest half no-stop does NOT beat live (0.71 vs 0.83) and without the top 2 it is ~0. Not verified; holdout fails. Matches a1-1993 (gain = 4 whales in first half).
Paper arm already exists at testmgr (stop.1+PULL+8h, from #2000); read forward n>=15 against the second-half band (~0.7x), not 2.5x. No new arm.