# #1922 bundle_all (296) x outcomes (analyst1, 11 Oct)
Script b218.py. Join bundle_all x e1-runners.csv: 185 of 296 have outcomes; e1-runners holds only pools that reached $300k, so every row is R300k=1. Result is CONDITIONAL on reaching 300k (the entry gate of the 300k-400k hunt), not a launch-time predictor. The 111 without outcomes never reached 300k (or not in e1).
Reach P1M / P3M given 300k, by bundle_usd quartile (all, n=185): <21.8k 49%/9% (n74); 21.8-28.4k 62%/12% (42); 28.4-35.6k 71%/32% (34); >=35.6k 88%/62% (32). Monotonic.
fee100 only (n79): 67/33 (6), 40/15 (20), 73/35 (26), 89/67 (27). Non-fee100 (n106): 47/7, 82/9, 62/25, 80/40 (n5): same direction, tiny hi bucket.
Out of sample: old_pre12sep (100 rows, none in my first 160): <35k 53%/6% (n95) vs >=35k 80%/40% (n5). ppdata window: 63/27 (n51) vs 86/62 (n29). Same sign in both.
Verdict: bundle_usd>=35k lifts P3M odds at the 300k gate roughly 2-3x; supports the BUNDLE35K arm (a1-bundle-size.md). Caveat: hi bucket outside fee100 is n5; no price-path/trade sim here. No new test sent: arm already specified; needs bundle_usd live (ppfeeds kit feed).
Gap: cluster/funder still NULL for new rows, so operator-prior lens cannot be rerun.