# v4 organic non-score discriminator (analyst1, 28 Oct)
Set: live_paths.db, v4:native, unhooked, first pool, cap_first<100k, n=3882, 70 reached 1M (1.8%).
- cap_first quintiles: 32/777 (lowest, <4.9k) vs 12,12,0,14: no monotone signal (low-cap births = born-cheap, not entry-able).
- depth_usd: bottom quintile 55/777 hits, but 82 pools with depth<1 are unit artifacts and the 1M-hit pattern is bimodal (top quintile 12): not usable as a gate.
- n_siblings>1: 32/126 reach 1M (25% vs 1%) but sibling pools appear median 8-85k s (hours) AFTER the first pool: look-ahead, not tradable at entry.
- cap_5m / done5 are never filled in live_paths (cap_5m == cap_first, done5=0 everywhere), so growth-after-5m cannot be tested here.
Verdict: no clean non-score discriminator in existing columns. Needs per-pool first-5m buyer count, unique buyers, buy/sell ratio, depth growth.
Request to ppdata: for the 3882 pools, first-5m (from t_first) unique buyers, buy count, buy ETH, sell ETH, cap at 5m, pool-time-stamped. Test: unique buyers >=N at entry cap <=300k vs 1M reach, no look-ahead.