Data sources

Props Oracle sources player box scores directly from official league data feeds. For the MLB launch, the primary source is mlb.com's official stats feed. Every league covered by the oracle has its primary source published in the per-league reference.

Before any assertion is posted, the primary source is cross-referenced against a second independent source. All stat values the oracle will assert must match exactly across both sources. Any mismatch blocks the assertion — the game is routed to manual review rather than posting with partial confidence. This does not guarantee correctness — both sources could theoretically agree on a wrong value — but it catches the substantial majority of data-entry and transcription errors, which are the realistic failure modes in practice.

The raw JSON responses from the MLB and ESPN sources, together with the play-by-play data used in the oracle's volatility assessment, are archived to Arweave before the assertion is posted. The Arweave transaction ID is included in the assertion payload, so disputers have a permanent, tamper-proof reference even if the original source is later edited or goes offline. The archive is the authoritative evidence record for what the relayer asserted and when; if a later official correction appears while the UMA window is still open, the relayer remains economically accountable through its bond.

What consumers can rely on

Two guarantees hold for every assertion. Every assertion references Arweave-archived source snapshots at posting time. Every assertion goes through a short formal UMA window after internal validation, described in Settlement → Challenge and disputes, during which any party can dispute and post a counter-bond. The data source strategy governs the baseline quality of asserted data; the dispute mechanism is the safety net if baseline quality fails.