Datasets and signals from prediction-market microstructure.
Dysrupt Labs operates Almanis, an operator-staked prediction-market platform in continuous operation since 2015. We extract three concurrent signals from the microstructure of the forecaster panel. Methodology peer-reviewed in the Journal of Financial Markets and eBioMedicine (The Lancet), replicated on independent DARPA programme data and, in March 2026, on a structurally different public forecasting venue.
The signals
Signal 1 — the crowd forecast.
Aggregated from the full forecaster panel. Tracks the public consensus benchmark for each release in a five-indicator US macro basket: CPI, NFP, GDP, PCE and Retail Sales.
Peer-reviewed: Journal of Financial Markets 2024.
Signal 2 — the separation between the crowd and the cohort.
A machine-learning-identified cohort within the panel produces forecasts whose accuracy advantage is regime-conditional. Signal 2 is the divergence between crowd consensus and cohort signal, expressed in the microstructure of the panel.
Peer-reviewed: eBioMedicine (The Lancet) 2023.
Signal 3 — scored divergence.
The magnitude of Signal 2, weighted by the cohort's historical track record on comparable episodes. The intent is to separate ordinary disagreement among forecasters from the structurally informative episodes that warrant attention. Methodology documented in a private whitepaper.
The same three-signal architecture can also be generated from the microstructure of liquid public prediction-market venues, as established by the March 2026 cross-platform replication. Available as a separate feed.
Provenance
The Almanis panel comprises approximately nine hundred active contributors, drawn from a candidate pool of more than thirty-six thousand, with a median tenure of seven years on the platform.
Why operator-staked
Almanis is operator-staked rather than player-staked. Forecasters are never required to put up money: they trade in virtual currency funded by Dysrupt Labs, and they are rewarded for forecasting performance — for demonstrated skill, not chance. There is no financial risk to participants and no counterparty risk between them.
The objective is to encourage and reward the speed and accuracy of forecaster contributions to the aggregate signals of the platform. This aligns the incentives of the forecasters with those of Dysrupt Labs and its clients.
Contact — contact@dysruptlabs.com