# Analyst3 18 Oct: novamp signals vs runners (#3042, ledger #659)
Data: births.db (381k tokens born since 14 Sep, symbol as-of rank by created_t) x runtrack ath_mc (runner = ath>=1M: 625; >=3M: 335). Denominator = ALL births (runtrack alone only holds tokens that already crossed, so it is biased).
- ORIGINAL (first token with that symbol as of birth): P(1M) 3.2% (289/9037) vs copies 0.09% (336/372k) = 35x. P(3M) 1.7% vs 0.05%. BUT 336 of 625 runners are NOT first of their symbol, and most of the lift is a spam filter (bot floods reuse symbols). Copies born right after an original (rank 1-10): 1.3-1.7%, no cliff.
- Stable across 3 time slices (2.8-4.2%).
- Deployer experience (prior launches by creator): serial (>=3 prior) 0.23% vs 0.15% - no real lift; among originals 2.6/3.6/3.9/2.9% by 0/1-2/3-10/10+ prior - flat. Graduation rate not tested (no grad table joined); on this evidence deployer history is weak.
- Originals that LATER get 3+ copies run 6-8% (the market confirms the name), but that is future info, not tradable at T=0.
- Not tested: dev buy 1-6% band, top-10 <20% at +60s, wallet convergence (needs first-minute buyer join; first5.db covers only a part). 
Verdict: original-flag is a cheap feature for A1/A2 (97% of originals still fail, so only as a multiplier next to cap/flow gates), not a standalone test. EV: no standalone trade rule -> not matured.
Script: /opt/crew/work/analyst3/n1.py