Dissmarket is committed to full methodological transparency. This page documents how we aggregate public opinion, source and normalise market odds, score predictions, and weight forecaster contributions. We update this page whenever we make material changes to our approach, and we maintain a changelog at the bottom so you can track how our methodology has evolved.
When a Dissmarket participant submits a prediction, they assign a probability to each outcome in a market. The crowd forecast is the aggregate of all individual predictions, weighted by each forecaster's historical accuracy.
Not all predictions are weighted equally. We use an accuracy-weighted aggregation that gives more influence to forecasters with stronger track records. The weighting formula considers three factors:
Brier score history. Forecasters with lower (better) aggregate Brier scores receive higher weight. This ensures that the crowd signal improves over time as better forecasters contribute more.
Category-specific performance. A forecaster who excels at political markets may not be equally skilled at technology markets. We apply category-specific weights so that each forecaster's influence is proportional to their demonstrated expertise in the relevant domain.
Recency. More recent predictions are weighted more heavily than older ones within the same market. This ensures that the crowd signal reflects the latest available information and does not anchor to stale estimates.
We also calculate and display an unweighted (simple average) crowd forecast for transparency. This allows users to compare the accuracy-weighted signal against the naive average and assess the impact of our weighting methodology.
A market must have a minimum number of participants before we display a crowd forecast. This threshold prevents small, unrepresentative samples from producing misleading signals. The specific threshold may vary by market type and is documented in each market's metadata.
Dissmarket aggregates odds from multiple prediction market platforms, betting exchanges, and bookmaker sources. Our primary data partners include regulated prediction markets, decentralised prediction platforms, and licensed bookmakers operating in relevant jurisdictions.
We source data through direct API integrations, licensed data feeds, and structured data collection from publicly available sources. All data sources are documented in our source registry, which is updated as we add or remove sources.