Research.
The signal suite rests on a peer-reviewed methodology. The publications below establish the underlying econometrics and the dataset's empirical credentials.
Peer-reviewed publications
Price Formation in Field Prediction Markets: the Wisdom in the Crowd.
Machine learning augmentation reduces prediction error in collective forecasting.
DOI: 10.1016/j.ebiom.2023.104783 · replicated on DARPA NGS2 programme data.
A consideration of group work processes in modern epidemiology.
Technical reports
Evaluation report on US macroeconomics.
Research notes
Signal behaviour under distributional instability.
Regime-dependent signal behaviour and execution implications. Draws on a working paper examining signal×uncertainty interaction across five uncertainty proxies.
Election forecast accuracy: cohort-signal performance across 174 markets.
ROC/AUC analysis of the cohort signal across 174 election markets (2017–2022) and nine US Presidential 2024 markets. Establishes the confirmation–divergence framework as a real-time signal-reliability indicator.
Cross-platform replication
In March 2026, the divergence and scored-divergence signals were replicated on a structurally different public forecasting venue using a constant-product automated market maker. The replication establishes that the signal is not specific to a single market-design or cost function and supports a behavioural rather than mechanism-design interpretation of its origin.
The dataset
The panel is generated by Almanis, our operator-staked prediction-market platform in continuous operation since 2015. The active forecaster population numbers approximately nine hundred contributors, drawn from a candidate pool of more than thirty-six thousand, with a median tenure of seven years on the platform.
Contact — contact@dysruptlabs.com