“Prediction markets excel at sports and elections but fail to inform most real-world decisions”
Dan Schwarz analyzes ~13,500 prediction market contracts from Polymarket to ask whether billion-dollar prediction markets deliver on their promise of informing decisions. Over 80% of volume goes to sports, crypto and elections, and accuracy on "useful" markets hasn't improved since early 2025. The piece argues AI chatbots may supersede prediction markets as the primary forecasting interface, leaving markets to serve an epistemic role as common knowledge infrastructure.
Some technical background helpful
Platforms mentioned: Polymarket, Kalshi, Metaculus, Manifold