The Polymarket Datasets Guide

Live dashboards read layer 3; filling-level backtests and audit work read layer 1; teaching and citable examples read layer 2; competitive scoping reads layer 4. The honest decision rule: match the layer to the claim — a paper citing "the dataset" must name which one.

The layers join cleanly because every timestamped one uses the UTC grid — a dashboard can overlay live and recorded panels without reconciliation.

This guide's job is legibility, not promotion: it names what exists, what each layer can prove, and where the recorded archive sits in that stack — and it points to the comparison pages for any claim about fitness for a specific use.

The guide's real value is traceability: a claim like "final-second books are single-sided 90% of the time" traces to layer 1 (recorded ladders), its teaching copy traces to layer 2 (open Zenodo), and a miss about live pricing traces to layer 3 (official feeds). Before trusting any Polymarket statistic, ask which layer produced it.

That four-layer discipline is what keeps this catalog honest about what each dataset can and cannot prove. The comparison pages extend the same rule to third-party sources.

• Open data
• Data sets page
• Archive