Dataset

polymarket-onchain-v1

A repository of roughly 2.7 million publicly available U.S. patent application records split into yearly source files spanning 2021 to 2026, covering application metadata, applicants and inventors, classifications, prosecution history, continuity and priority data, assignments, publications, and grant information where present, sourced from the USPTO.

Open on Hugging Face

Data preview

Subset

Loading preview...

Dataset card

The Polymarket On-Chain v1 dataset provides a comprehensive, queryable record of every trade, position lifecycle event, and payout on the Polymarket prediction market from its 2020 launch through April 2026. Hosted by moose-code and licensed CC-BY-4.0, it spans billions of records covering over a billion on-chain trades and $59.9B in volume.

What this dataset contains

The dataset captures the full on-chain lifecycle of Polymarket activity on the Polygon blockchain, extending beyond simple trade records. It includes every CLOB fill as well as conditional-token split, merge, redemption, and resolution events, enabling reconstruction of real positions, realized PnL, and resolution payouts. The training split alone contains 2,734,403,098 rows.

Structure and fields

The schema includes the fields id, positionIds, payoutNumerators, and payoutDenominator, all serialized as string values. Underlying tables such as order_filled are partitioned by time, allowing tools like DuckDB to scan only the byte ranges needed for a query without downloading the full file. The data is accessible over HTTPS via paths such as hf://datasets/moose-code/polymarket-onchain-v1/order_filled//*.parquet.

How analysts use it

Analysts can run SQL directly against the partitioned Parquet files to study trade activity, market liquidity, and resolution outcomes across the platform's history. The open-source indexer code built on Envio HyperIndex allows every aggregate to be reproduced from raw Polygon data. Fundamentally, the dataset supports research into prediction-market pricing, trader concentration, and realized PnL distributions at a granular level.

Documentation & references 1