An alternative data feed that detects when informed forecasters reliably disagree with market consensus.
We operate a private operator-staked prediction market of 900+ forecasters on US macroeconomic releases and other variables of interest. Our machine learning generates three concurrent signals. Approximately 95% of the time, the signals confirm each other. In the rare episodes when they diverge, the general consensus has frequently revised toward the insight estimate.
The crowd forecast. Aggregated from 900+ forecasters with a 7+ year median tenure. Well-calibrated in normal conditions.
ML-filtered signal that amplifies forecasters identified as disproportionately accurate in real time. The edge in the roughly 5% of divergent episodes.
Z-scored distance between the two signals. When it spikes, the consensus typically revises toward the insight estimate—before the print.
The insight signal methodology is published in the Journal of Financial Markets (2024) and the ML augmentation validated in eBioMedicine (2023), both replicated on independent DARPA programme data. In aggregate across 174 election markets, insight and general signals perform similarly (AUC 0.87 vs 0.86; eBioMedicine, 2023). The insight advantage concentrates in the rare episodes when the two diverge. Current research examines signal behaviour under distributional instability and forecast accuracy across 174 election markets.
A continuous minute-level data feed of three signals: general consensus, ML-filtered insight estimate, and divergence z-score. Most of the time the two estimates agree, confirming the crowd is well-calibrated. In the roughly 5% of episodes where they diverge beyond threshold, the insight signal carries a measurable accuracy advantage and the general consensus typically revises toward it. The system captures this drift—it does not depend on how the underlying question resolves.
Coverage: CPI, NFP, GDP, PCE, Retail Sales, and Housing Index. Delivery via REST API, WebSocket, or SFTP.
Product detail →Founded by Karl Mattingly (Columbia MBA, 25 years institutional finance) with Professor Anne-Louise Ponsonby (500+ peer-reviewed publications, Florey Institute) as Chief Scientific Adviser. Platform architecture led by Chad Nash (PhD Quantum Physics, 10-year tenure).
Full team →Past performance is based on backtested data and is not indicative of future results.