# Preliminary

## Preliminary Financial Time Series Dataset

> ⚠️ **Warning:** This dataset is preliminary and intended for development, testing, and feedback purposes only.  
> A stable, production-ready version will be released later as a separate dataset.

## Overview

This dataset contains **parquet files** with time series data for various financial instruments, collected through an early-stage version of our data pipeline. The data is already usable, but the structure and naming conventions are **still subject to change**.

Each instrument currently has:
- One file with **daily candles** covering multiple years
- One file with **minute candles** covering a much shorter, but more detailed recent period

This dual-resolution format is intended to enable models that can reason across both long-term and short-term patterns.

## Purpose

This preliminary dataset exists to:
- Make early access to usable data possible
- Facilitate testing and iteration
- Gather feedback before finalizing a stable format

## Expected Changes

The internal structure of the parquet files will remain stable. However, the following may change:
- File naming conventions
- Organization into folders or shards (e.g., by year or month)
- Dataset split logic

The final version will be released separately to ensure compatibility and stability for users who rely on long-term consistency.

## Goals
Datasets in Traders-lab will continue to grow in two aspects:
* more stocks  
    More stocks will be available here over time.  Expect the number of available stocks to grow rapidly in the near future.
* More data, especially in short-term data sets (minutes and hours)
    The time data will accumulate over time, so the available time spans for the minute and hour data sets will grow soon.  
    I cannot yet promise that the currently available minute data will certainly be included in a complete time history, but in the near future I will be able to ensure this continuity. This means that the available time data will then be continuously updated and will grow steadily.

Translated with www.DeepL.com/Translator (free version)

## Non-Goals
The data should be sufficiently up-to-date for training purposes. 

It is not a goal to be up to date.
Up-to-date enough for training purposes means that the data is at most a few days old, but several days can and may occur before an update.

### License & Usage

This dataset is not licensed under a standard open data license.

See the [`LICENSE`](LICENSE) file for usage permissions. The dataset is intended for **solely for research and educational purposes**.  
Data is aggregated from publicly available sources such as [Yahoo Finance](https://finance.yahoo.com) via `yfinance`.

Redistribution may be restricted. Please respect the terms of the original data providers.