Ranked by consistency and copyability, not size. Green = days profitable out of the 7-day equity curve; 1-day conc = share of the week's gains earned on its best single day (high = one lucky event). Measured 260821: this ranking and the raw-PnL ranking overlap 0/5 on SOL, BSC and ETH.

Every graded wallet's position on the graduation path, and the first criterion blocking it. Promotion is earned on resolved paper outcomes, never on the source wallet's own PnL — magnitude is what wash volume and position size inflate.

GRADUATING TO S — all seven must hold
1. no copy-disqualifying nature (WSH/INST/MM/BOT)
2. ≥20 resolved paper copy decisions
3. positive expectancy vs matched control, temporal OOS, FDR-controlled
4. green_ratio ≥ 0.7 across ≥2 independent 7d windows
5. single-day concentration ≤ 0.5, regime-adjusted to its chain
6. avg_cost inside our mirrorable band (≤$3,000)

S IS NOT PERMANENT — demote on any of
· nature gains a copy-disqualifier (WSH/INST/MM/BOT)
· 2 consecutive windows failing consistency or concentration
· resolved expectancy turns negative vs matched control

FLAGS (noted, never gating)
· crowded — audience is large; the trade may already be public

LADDER 🌱CANDIDATE → ⏳PROBATION → 🟣S → ⚠️REVIEW → ⬇️DEMOTED

Promotion is earned on RESOLVED PAPER OUTCOMES, never on the source wallet's
own PnL — magnitude is the metric wash volume and position size inflate.

Wallets we deliberately never copy — institutions and market makers trade size we cannot mirror; wash traders and 66k-follower influencers put us last in the queue. Their presence is still evidence, so it reaches the stack as a token-card boost and as front-run seeds instead of being discarded.

Where discovered alpha goes, and where it stops. Nothing is deleted — OBSERVED wallets stay in the DB and simply drive nothing yet.

OBSERVED 2,119 · SMART_MONEY_BOOST 121 · EARLY_SNIFFER 120 · COPY 75 · SELECTIVE_ENTRY 15

COPY — consistent, uncrowded, our position size → paper copy slots
SELECTIVE_ENTRY — rare precise entries → its own bucket, measured separately
SMART_MONEY_BOOST — institution / MM / VC — cannot copy, but presence is evidence → card boost
EARLY_SNIFFER — wash / high-frequency — cannot copy, but touches tokens early → discovery
FRONTRUN_TARGET — crowded + profitable → find who buys BEFORE it
OBSERVED — held in the DB, drives nothing yet

Copy score against follower count. Anything right of the gold line is a trade the market already knows. Top-left — high score, few followers — is the uncrowded edge.