alexmindustry/cryptotradr-research-data

Cryptotradr Market and Research Data

This dataset preserves the market-data and research-artifact portion of the cryptotradr project while retaining the original directory structure.

Contents

• kline_data/: full historical cryptocurrency kline/candlestick collection.
• kline_data_sample/: the project's original sample collection, intentionally preserved even where it overlaps the full collection.
• *.pkl: all Python pickle research artifacts found throughout the original project tree, including historical DataFrame batches, backtest/plot bundles, parameter-search outputs, event windows, ML arrays, and symbol metadata.
• *.npz: all NumPy NPZ artifacts found during publication. The publication preflight found no NPZ files at the time of upload.
• manifest.csv: path, byte size, SHA-256 digest, and extension for every uploaded source artifact.
• SHA256SUMS: integrity digests in standard checksum format.
• DATA_STRUCTURE.md: schemas and pickle-family descriptions.

At publication preflight, the selected source inventory contained 32,770 files totaling approximately 78.434 GiB: 26,719 CSV files and 6,051 pickle files.

Kline CSV schema

The kline files use the following columns:

1. open_time
2. open
3. high
4. low
5. close
6. volume
7. close_time
8. quote_asset_volume
9. number_of_trades
10. taker_buy_base_asset_volume
11. taker_buy_quote_asset_volume
12. ignore

Files are partitioned primarily as:

kline_data/<SYMBOL>/<SYMBOL>_<YEAR>_<INTERVAL>.csv

The collection covers 819 trading pairs, years 2018 through 2023, and intervals including 1m, 3m, 5m, 15m, 30m, 1h, 2h, 3h, 4h, 6h, 8h, 12h, 1d, 3d, 1w, and 1Month.

Pickle safety and compatibility

The pickle files are original Python research artifacts created by the dataset publisher. Python pickle is executable serialization: do not deserialize these files unless you trust this repository and your Python environment. They may depend on historical versions of Python, pandas, NumPy, scikit-learn, or related libraries.

Pickles are retained unchanged for reproducibility and archival fidelity. See DATA_STRUCTURE.md for the known producer-side structures.

Provenance

The kline schema and project collection code indicate that the market data was collected from Binance APIs for cryptocurrency research. The included pickle files are derived research artifacts generated by the cryptotradr notebooks and scripts.

Associated source-code repository: https://github.com/alexpvpmindustry/cryptotradr

Duplication

kline_data_sample/ is intentionally included as requested by the publisher. Most of it overlaps kline_data/; consumers who only need unique market records should normally use kline_data/.

License and third-party terms

The dataset metadata uses license: other. The project source code is GPL-3.0, but that software license does not automatically grant rights in third-party market data. Binance-derived records remain subject to applicable Binance terms, attribution requirements, and other legal restrictions. Users are responsible for confirming that their intended use and redistribution comply with those terms.

Research artifacts authored by the publisher are provided for research and reproducibility without warranties. No ownership claim is made over third-party market data.

Limitations

• Coverage is not guaranteed to be complete for every symbol, year, or interval.
• Newly listed or delisted pairs naturally have shorter histories.
• Some sample files overlap the full collection.
• Pickle files are Python-specific and not directly supported by the Hugging Face Dataset Viewer.
• The research artifacts reflect experimental strategies and may contain intermediate or superseded results.

Financial-use disclaimer

This dataset is provided for research and educational purposes. It is not financial advice, and no trading performance, correctness, completeness, or fitness for a particular purpose is guaranteed.

Integrity

Use manifest.csv or SHA256SUMS to validate downloaded files. Paths in the manifest are relative to the dataset root and preserve their locations from the original cryptotradr project.