Bedrock Market Data Archive

Bedrock is a free and open collection of historical market data covering equities, cryptocurrencies, and other assets across a range of timeframes. The project is built for quantitative researchers, machine learning practitioners, backtesters, students, and developers who need accessible historical data without having to pay for another data API or piece together files from different sources.

The goal is straightforward: keep useful historical market data in one place, make it easy to inspect and download, and keep the underlying files simple enough that they can be used with almost any research stack.

Why this exists

Reliable historical market data is often harder to obtain than it should be. Some datasets are locked behind paid APIs, while others are spread across different providers and stored using inconsistent conventions. Even when the data is available, getting several assets and timeframes into a form that is actually convenient for research can become a project of its own.

Bedrock is an attempt to solve that problem with a simple archive of openly accessible historical data. The project focuses on practical usability rather than building another complicated data platform. The files remain available in plain CSV format so that they can be inspected directly, processed with standard tools, or incorporated into larger research pipelines.

The archive is intended to grow over time as new assets, timeframes, longer historical periods, corrections, and metadata are contributed.

Repository structure

Data is organized by ticker, with the timeframe represented in the filename.

datasets/
├── AAPL/
│   ├── AAPL_2025.csv
│   └── AAPL_USD.csv
├── BTC/
│   └── BTCUSD_1-MIN_DATA.csv
├── ETH/
│   ├── ETH_1H.csv
│   ├── ETH_1min.csv
│   └── ETH_day.csv
└── SOL/
    ├── SOLUSDT_1minute.csv
    ├── SOLUSDT_5minutes.csv
    ├── SOLUSDT_15minutes.csv
    ├── SOLUSDT_30minutes.csv
    ├── SOLUSDT_1hour.csv
    ├── SOLUSDT_2hours.csv
    ├── SOLUSDT_3minutes.csv
    ├── SOLUSDT_4hours.csv
    ├── SOLUSDT_6hours.csv
    ├── SOLUSDT_8hours.csv
    ├── SOLUSDT_12hours.csv
    ├── SOLUSDT_1day.csv
    ├── SOLUSDT_3days.csv
    ├── SOLUSDT_1week.csv
    └── SOLUSDT_1month.csv

Each ticker has its own directory. File names generally describe the asset and candle resolution, such as 1minute, 1hour, 1day, 1week, or 1month.

Some files retain naming conventions from their original sources. This means that naming and column conventions are not yet perfectly uniform across the archive. Standardization is an ongoing part of the project rather than something being hidden behind a single artificial schema.

Data format

The datasets are primarily provided as standard CSV files containing OHLCV style market data.

A typical file contains the following fields:

Column	Description
timestamp / date	Candle opening time
open	Opening price
high	Highest price during the period
low	Lowest price during the period
close	Closing price
volume	Traded volume

The exact column names and available fields can differ between files because the archive currently contains data originating from different sources. Timestamps are generally represented in UTC where that information is known, but users should check the individual dataset before assuming a particular timezone or convention.

If you find inconsistent naming, malformed data, missing information, or another issue, contributions and corrections are welcome.

Current coverage

Bedrock currently contains historical data for AAPL, BTC, ETH, and SOL.

AAPL contains historical Apple equity data. BTC contains BTCUSD data at 1 minute resolution. ETH contains Ethereum data across 1 minute, 1 hour, and daily resolutions.

SOL currently has the broadest timeframe coverage in the archive, ranging from 1 minute through 1 month. This includes 1 minute, 3 minute, 5 minute, 15 minute, 30 minute, 1 hour, 2 hour, 4 hour, 6 hour, 8 hour, 12 hour, 1 day, 3 day, 1 week, and 1 month candles.

Coverage will change as the archive grows, so the repository itself should be treated as the current source of truth for available assets and files.

Using the data

The files can be downloaded directly from Hugging Face or accessed through the repository using your preferred tooling.

For a local copy of the repository, clone it with:

git clone https://huggingface.co/datasets/akhverm/Bedrock

A CSV can then be loaded directly with pandas:

import pandas as pd

df = pd.read_csv(
    "datasets/SOL/SOLUSDT_1hour.csv"
)

print(df.head())

You can also load an individual file directly from Hugging Face without cloning the entire repository:

import pandas as pd

url = "https://huggingface.co/datasets/akhverm/Bedrock/resolve/main/datasets/SOL/SOLUSDT_1hour.csv"

df = pd.read_csv(url)

print(df.head())

The same files can be used with Polars, DuckDB, PyArrow, or other tools capable of reading CSV data.

Time series research

Bedrock is intended to be useful for research involving historical financial time series. Once a dataset has been loaded, users can build features, calculate returns, construct signals, train forecasting models, perform statistical analysis, or use the data as an input to a backtesting framework.

For example:

import pandas as pd

df = pd.read_csv(
    "datasets/SOL/SOLUSDT_1hour.csv"
)

df["timestamp"] = pd.to_datetime(df["timestamp"])
df = df.sort_values("timestamp")

df["return"] = df["close"].pct_change()

print(df.head())

When using the data for machine learning or backtesting, care should be taken to preserve the temporal structure of the dataset. Randomly shuffling financial observations can introduce look ahead bias and produce results that do not represent how a strategy or model would have behaved historically.

Researchers should also account for issues such as missing observations, duplicate timestamps, market closures, exchange outages, corporate actions, survivorship bias, transaction costs, and slippage where relevant to their experiment.

Data provenance

Bedrock is an evolving archive and may contain data collected or derived from multiple sources. Different providers can use different conventions for timestamps, symbols, volume, market sessions, and historical adjustments.

For that reason, users should verify the provenance and characteristics of an individual file before relying on it for a particular research project. The intention is to improve source documentation and metadata as the archive develops.

The project does not claim that every file is perfectly cleaned or standardized. Transparency about the underlying data is considered more useful than presenting an apparently uniform dataset without documenting how that uniformity was produced.

Disclaimer

The data in Bedrock is provided as is for research, educational, and experimental purposes.

No guarantee is made regarding the accuracy, completeness, continuity, or fitness of the data for any particular purpose. Historical market data can contain errors, missing observations, unexpected gaps, or source specific inconsistencies.

Bedrock should not be treated as a source for live trading decisions or financial advice. Data that is important to a financial decision should be independently verified against an appropriate primary or authoritative source.

Contributing

Bedrock is intended to be a community maintained archive. New assets, additional timeframes, longer historical coverage, data quality fixes, metadata, documentation improvements, and useful tooling are all welcome.

If you find a problem with an existing file, please document the issue clearly and open a contribution where possible. If you are adding new data, include whatever information is available about its source, timeframe, timezone, and methodology.

Contribution guidelines are available in CONTRIBUTING.md.

License

Bedrock is released under the MIT License.

See LICENSE for the complete license text.

The licensing of the underlying market data may depend on its original source. Users are responsible for checking applicable source specific terms where relevant.

Maintainer

Bedrock is maintained by github.com/ak495867.

The project is built around a simple idea: historical market data should be easier to access, inspect, and experiment with. If you use the archive in a project, research workflow, or backtesting system, contributions and improvements are welcome.