Ashitemaru/Quant-Arena-Public-L1

Quant-Arena Public L1

Quant-Arena Public L1 is the public market-data release used by the Quant-Arena benchmark. It contains Level-2 order-book quotes (ten-level bid/ask snapshots) for a fixed universe of 289 China A-share stocks, resampled to a regular 15-second grid for every trading day from 2022-01-01 to 2023-12-31 (484 trading days). No data from 2024 or later is included.

• Stocks: 289 (a seeded 10% sample of 2,894 A-share stocks with codes 00xxxx and 30xxxx)
• Rows: 131,889,757
• Columns: 45 (symbol, available_at, source_event_time, ten levels of bid/ask price and volume, missing_depth, stale)
• Format: Parquet (zstd), one file per stock, about 6.23 GB

Values are the exchange's published snapshots. No order book is rebuilt and nothing is interpolated or adjusted. Each row uses only information available at its available_at timestamp. The full data card (schema, sampling rule, null conventions, flags) is README.md inside the archive, and per-file row counts, sizes, hashes and date coverage are in manifest.json.

Files in this repository

File	Content
quant-arena-public-l1-v1.tar	Plain tar archive of the release directory quant-arena-public-l1-v1/
quant-arena-public-l1-v1.tar.sha256	SHA-256 of the archive
README.md	This card

Inside the archive:

quant-arena-public-l1-v1/
  <symbol>.parquet   289 files, e.g. 000012.SZ.parquet
  manifest.json      release metadata, schema, per-file rows/bytes/sha256/date range
  SHA256SUMS         checksums of every file in the directory
  README.md          full data card
  LICENSE            CC BY-NC 4.0 legal text

Download and verify

huggingface-cli download Ashitemaru/Quant-Arena-Public-L1 --repo-type dataset --local-dir quant-arena-l1
cd quant-arena-l1
sha256sum -c quant-arena-public-l1-v1.tar.sha256
tar -xf quant-arena-public-l1-v1.tar -C .
cd quant-arena-public-l1-v1 && sha256sum -c SHA256SUMS

(Equivalently, from the parent directory: tar -xf quant-arena-l1/quant-arena-public-l1-v1.tar -C quant-arena-l1.)

Reading the data

import polars as pl
df = pl.scan_parquet("quant-arena-l1/quant-arena-public-l1-v1/*.parquet").collect()

Prices are decimal128(28, 4) in CNY per share; volumes are int64 shares; timestamps are Asia/Shanghai (UTC+08:00). An empty book level is null in both its price and volume columns.

Licence

Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). See https://creativecommons.org/licenses/by-nc/4.0/legalcode.

Disclaimer

This dataset is provided for research and educational purposes only, "as is", without warranty of any kind. Nothing in it is investment advice.

Citation

If you use this dataset, please cite:

@misc{quantarena2026,
  title  = {Quant-Arena: Benchmarking LLM Agents for Quantitative Trading Research},
  author = {TBD},
  year   = {2026},
  note   = {Citation details to be added}
}