We extracted 1,501 predictions from 129 episodes of the All-In Podcast and scored them against what actually happened. Not just right or wrong — precision AND directional accuracy.
AI processes full episode transcripts to identify prediction-like statements. Speaker detection assigns each call to the right Bestie.
Each prediction is verified against real-world outcomes using web research and cross-validated with multiple sources for accuracy.
Weekly pipeline auto-detects new episodes, extracts fresh predictions, and updates the scorecard. Pending predictions get scored as outcomes unfold.
Get notified when new predictions are extracted, scored, and when the leaderboard changes.