Peer-Reviewed Publications
Journal of Financial Markets, 2024
Price Formation in Field Prediction Markets: the Wisdom in the Crowd
Documents the algorithm that separates informed trades from noise in a prediction market, generating the insight signal. Approximately 7% of trades are classified as price-sensitive, providing the algorithmic basis for the ML accuracy weighting that produces the insight estimate delivered alongside the general consensus in the commercial data feed.
eBioMedicine / The Lancet, 2023
Machine learning augmentation reduces prediction error in collective forecasting
Demonstrates that ML accuracy weighting reduces forecast error by 18–23% versus equal-weight averaging. Overall AUCs are similar; the insight advantage concentrates in the minority of episodes where the two signals diverge. Replicated on DARPA NGS2 programme data.
Annals of Epidemiology, 2014
A consideration of group work processes in modern epidemiology
Foundational framework for group-level inference under uncertainty. Establishes the epistemological basis for the collective intelligence architecture underlying the platform.
Technical Reports
Dysrupt Labs, 2025
Evaluation Report on US Macroeconomics
39-month backtest (March 2019 – June 2022) of the hybrid alert system across 7 FX pairs and 51 macro-sensitive ETFs. 4,710 trades, two exit strategies (trailing stop-loss and 540-minute time-based). The ETF backtest generated portfolio alpha of 697 bps (time-based exit) and 295 bps (optimized trailing stop loss) annualised, both relative to BIL. Includes CAPM-based evaluation, simulation quasi-experiments, and causal inference assessment.
Research Notes
Signal Behaviour Under Distributional Instability
Regime-dependent signal behaviour, quintile inversion under Knightian uncertainty, Kyle lambda regime shift, and implications for execution design. Draws on current working paper examining signal×uncertainty interaction across five uncertainty proxies.
Election Forecast Accuracy: Insight Signal Performance Across 174 Markets
ROC/AUC analysis of insight vs. market signals across 174 election markets (2017–2022) and 9 US Presidential 2024 markets. Establishes the confirmation–divergence framework as a real-time signal reliability indicator. Insight AUC 0.87 vs. market 0.86 across 174 election markets; advantage concentrates in high-divergence regimes.