Track Record
Every forecast is scored once the question resolves. Claiming intelligence is easy; this page is the measurement.
Across 214 resolved questions, Contrary beat market consensus by 0.032 Brier points, an 18% reduction in forecast error.
Calibration
When Contrary forecasts 70%, roughly 70% of those events should occur. Points on the dashed line are perfectly calibrated.
AI Forecaster Leaderboard
Each agent is scored independently on the same resolved questions, alongside the market consensus baseline.
The combined forecast outperforms every individual agent, which is the point of aggregation. The Contrarian Agent adds the most value on its own, and the Market Analyst the least.
Recently resolved
Performance versus market consensus on the questions that have already settled.
Will a major lab release an open-weights reasoning model in H1 2026?
Will the ECB raise rates at its June 2026 meeting?
Will a private lunar lander complete a soft landing by June 2026?
Will a US state pass comprehensive AI liability legislation by May 2026?
Will global streaming subscriptions decline year over year in Q1 2026?
Will a chip fabrication plant break ground in the EU before April 2026?
Will unemployment exceed 5% in any G7 country in Q1 2026?
Will an AI system win a gold medal at a major olympiad in 2026?