elthariel/histdata_fx_1m

HistData FX 1m

This is an export in the parquet format of the histdata.com dataset, which contains mostly Forex market tickers data.

For each symbol, 2 files are provided:

• ticks.parquet contains the ticker (ohlc) data
• gaps.parquet contains the reported time gaps

The timestamps are converted from the original EST timezone to UTC.

Ticks

The ticks.parquet files have the following schema:

{
    'ts': pl.Datetime,
    'open': pl.Float64,
    'high': pl.Float64,
    'low': pl.Float64,
    'close': pl.Float64,
    'volume': pl.UInt64,
}

Gaps

The gaps.parquet files have the following schema:

{
    # The gap duration
    'length': pl.UInt64,
    # The timestamp of the gap start converted to UTC.
    'start': pl.Datetime,
    # The timestamp of the gap end converted to UTC.
    'end': pl.Datetime,
}

Original dataset.

The original dataset can be found at HistData.com and the data was downloaded using this tool

Updating the dataset

The dataset can be updated using tools from the following repo: https://github.com/elthariel/FX-1-Minute-Data

FX_DATA_OUTPUT=path_to_this_repo python path/to/download_all_fx_data.py

The data of the current year is downloaded per-month, but the older years are downloaded per-year. So you'll need to clean some extra files if you run the script in a new year, with something like: rm */DAT_ASCII_*_M1_2024??.zip

Then you need to process the individual files into the concatenated per-symbol datasets

cd path_to_this_repo && python datafux.py extract .

License

The license is unclear. histdata.com has the copyright and provides the data for free, but do not explicitly provide a license for it. So I think this is OK to upload this, and for you to download it and train stuff on it, but this isn't a certainty and I provide no legal guarantees whatsoever.

If you're the owner of histdata.com and would like me to remove this data. Please reach out and I'll oblige.