polymarket_historical_data

Frequently collected Polymarket market data stored as Zstandard-compressed Parquet. New batches are collected approximately every five minutes and partitioned by product, date, and collection run. Public upload of collected data began on July 23rd, 2026.

Data Products

• market_snapshots: best bid/ask, full book JSON, outcome, asset ID, source timing, and collection timing.
• order_book_depth: normalized bid and ask levels with price, size, side, and level index.
• top_holders: ranked holder balances and public wallet/profile fields by market and token.

Files use this layout:

data/<product>/date=YYYY-MM-DD/<run-id>.parquet

Usage

from datasets import load_dataset

snapshots = load_dataset(
    "trentmkelly/polymarket_historical_data",
    "market_snapshots",
    split="train",
)

Parquet files can also be queried directly with DuckDB, Polars, PyArrow, or other compatible tools.

Citation

@misc{kelly_polymarket_historical_data_2026,
  author       = {Trent Kelly},
  title        = {Polymarket Historical Data},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/trentmkelly/polymarket_historical_data}}
}

License

This dataset is made available under the Creative Commons Attribution 4.0 International license. Commercial use and adaptation are permitted with attribution. This license applies only to rights the dataset maintainer is authorized to license.

Notes

• Timestamps are UTC ISO 8601 strings.
• Coverage prioritizes active markets with higher recent volume and may change as markets open, close, or become unavailable.
• Empty or unavailable API responses can produce gaps; do not assume every market appears in every run.
• This is an independently collected research dataset and is not an official Polymarket publication. Users are responsible for complying with applicable terms and regulations.