Research

CNBC pundit pick alpha, measured.

Every stock pick from Halftime Report and Fast Money, transcribed, scored against SPY across 30/90/180-day windows. Updated as new clips air. The leaderboard ranks by conviction-weighted 90-day alpha — your highest-conviction pick counts more than your throwaway flier.

Subscribe to the Substack for new findings.

Leaderboard

top 12 · min 5 picks · conv-weighted 90d alpha

#PunditPicksHit rate (90d)Avg α (90d)Conv-wtd α (90d)
1Jenny Harrington7100.0%+17.08%+15.31%
2Unknown64100.0%+13.28%+13.16%
3Joe Terranova5075.0%+10.00%+13.08%
4Farmer Jim1350.0%+4.63%+6.42%
5Kevin Simpson3358.3%+5.81%+5.43%
6Jenny2671.4%+4.64%+5.24%
7Tim Seymour5261.5%+3.94%+4.83%
8Jim1066.7%+6.33%+3.75%
9Dan1350.0%+2.93%+2.93%
10Sarat Sethi3744.4%+0.58%+2.20%
11Jim Lebenthal7100.0%+2.09%+2.09%
12Stephanie Link5842.3%+1.75%+1.67%

32 more below the cut. See the full leaderboard →

Methodology

Source: every Halftime Report and Fast Money clip from CNBC's sitemap. Each is transcribed (Whisper), and an LLM extracts individual stock picks tagged with commentator, ticker, direction (long/short/avoid), conviction (1–5), and rationale.

Returns: forward 1-, 30-, 90-, 180-, and 1-year price changes from pick date, alpha calculated against SPY over the same window. Short picks have their sign flipped. Picks the user marks hidden in Notion are excluded.

Names: Whisper mishears ("Belski" → Brian Belsky, "Sechan" → Rob Sechan) are corrected via the Pundit Watchlist in Notion. Aggregations join on the canonical name, not the raw transcript token.

Commentary score: a separate LLM pass rates each panelist's clarity, falsifiability, and originality. The interesting tension is when the two diverge — pundits who reason rigorously but whose picks don't make money, or vice versa.