Oddpool is probability infrastructure for prediction markets. We capture full orderbook microstructure across venues and normalize it into structured, queryable data for quantitative and institutional workflows. Parquet-native, with full provenance. A Federal Reserve research paper found that prediction markets provide a statistically significant improvement over traditional forecasts for major macro events. But the data is fragmented across venues, unstandardized, and has no institutional-grade historical record. Oddpool provides the normalized cross-venue data layer, the canonical event mapping that links the same outcome across exchanges, and the historical depth that compounds over time and can't be rebuilt retroactively.