KOL Leaderboard

Whose voice gets weight, and why. Tier-1 = the four hand-curated names always trusted as an anchor. Tier-3 = active asserters earned in by raw posting density. Tier-2 — the promotion zone — needs price-outcome scoring to gate handles in, and is empty until events_raw is unblocked.

How to read: Assertion counts come from the extractor pass over kol_posts_raw. Stance mix shows what the handle has been pushing (bullish / bearish / paid-shill). Empty rows for tier-1 anchors mean none of the four have posted since the whitelist was seeded — present for credit-assignment, not for activity.

The readNo measured migrated callsthin: 0 measured

No handle has a measured migrated call yet, so KOL entry-timing quality is unproven on this sample. KOL tiers tell you whose voice gets structural weight in the convergence signal — NOT which calls to copy-trade (pre-migration mention rate ~1% of posts; the copy-trade backtest shows noise, not edge). KOL mentions LAG; creator-reputation t=0 is the only validated leading signal.

⚠ A handle with high assertion count but a low Pre-Mig % is a cheerleader calling tops. Tier ≠ timing quality.

→ Sort the Alpha Scores table by Pre-Mig % descending. Handles with >50% before-migration rate called the move, not the wick.

0 KOL profile(s)

Sorted by tier, then by assertion count. Tier-1 anchors first regardless of activity.

Alpha scores — token-selection quality

reverse_discovery loop: did this handle's resolved-mint calls land on tokens that are still alive at $500k+ MC? Pumped 10x = MC >= $500k (10× pump.fun baseline). PT cols = paper_trades_simulated ladder ROI (v0_2x50/v1_15x33). Note: this measures token-selection — NOT entry timing. A high alpha rate + low PT ROI means 'picks winners, but tweets too late to trade'.

Copy-trade wallets (180d)

PSX-Pulse salvage. Each KOL's calls mirrored into a simulated equal-size wallet from the per-call blended-ladder return (paper_trades_simulated.realized_pct) — enter at the call, marked to the ladder exit. SIMULATED, not real fills. Sorted by MEDIAN return (robust: memecoin OHLCV throws fat-tail spikes — a single thin-liquidity candle can print +247616%, so a mean lets one tick crown a wallet). Mean kept as a second column so the upside tail stays visible. Ranked only at ≥3 priceable calls; below that = warming up (no score). is_control + look-ahead filtered out.

Conviction lexicon (60d)

Per-handle stance averages from kol_post_stance (CryptoBERT). Permabulls = handles whose posts skew bullish (high bullish_ratio = most posts top-labeled bullish). Skeptics = bearish lean. Pairs with hit-rate below: a permabull with 0% hit-rate is a cheerleader; a skeptic that's right is a contrarian. Joined with kol_cohorts for tier × cohort context.

Cheap hit-rate (60d)

PSX-Pulse-inspired complement to paper-trade EV: classifies each paper_trades_simulated row by exit_reason (hit / miss / expired / unmeasured), aggregates per-handle + per-cohort. Different lens from ctrl_deep bootstrap CIs — that asks 'does the cohort beat random shuffles?', this asks 'which individual handles ever scored?'. Cohort roll-up is the sharpest view (a class of voices' average behavior).

Recent convergence events

kol_convergence_events — patterns: convergence, divergence, echo_chamber, kol_overlap (relaxed fallback while tier-2 is offline). Newest first.