klines-bybit / README.md
rogerdehe's picture
initial dataset card (under construction)
f571ee6 verified
|
Raw
History Blame Contribute Delete
2.1 kB
---
license: mit
task_categories:
- time-series-forecasting
language:
- en
tags:
- crypto
- cryptocurrency
- finance
- ohlcv
- klines
- bybit
- futures
- perpetual
- trading
pretty_name: Bybit USDT Perpetual OHLCV Klines
size_categories:
- 1B<n<10B
---
# Bybit USDT Perpetual OHLCV Klines
Historical OHLCV candlestick (kline) data for **Bybit** USDT-margined **perpetual
futures**, stored as zstd-compressed Parquet. Includes **delisted symbols**, so it
is usable for point-in-time / survivorship-bias-free backtests.
> ⚠️ **Under construction.** This dataset is being actively built. Coarser
> timeframes (15m · 1h · 4h · 1d) and 5m/3m are being uploaded first; **1m is
> still downloading** and will be added. Pair and timeframe coverage will grow.
> A per-file `COVERAGE.md` will be published once the build settles.
## Coverage (in progress)
| | futures (USDT perp) |
|---|---|
| **Pairs** | ~617 (incl. delisted) |
| **Timeframes** | 15m · 1h · 4h · 1d · 5m · 3m present; **1m coming** |
| **History** | from each contract's listing (mostly 2020–2021 onward) → present |
| **Extras** | `mark` price and `funding_rate` series for most pairs |
## File layout
```
futures/{timeframe}/{PAIR}.parquet
```
Examples:
- `futures/1d/BTC_USDT_USDT.parquet`
- `futures/1h/BTC_USDT_USDT-mark.parquet` (mark price)
- `futures/8h/BTC_USDT_USDT-funding_rate.parquet` (funding rate)
Pairs are `BASE_USDT_USDT` (USDT-margined perpetual). Each OHLCV Parquet has
columns `date` (UTC candle open), `open`, `high`, `low`, `close`, `volume`.
## Usage
```python
from huggingface_hub import hf_hub_download
import pandas as pd
path = hf_hub_download("rogerdehe/klines-bybit",
"futures/1d/BTC_USDT_USDT.parquet", repo_type="dataset")
df = pd.read_parquet(path)
```
## Notes
- Timestamps are UTC. Coverage varies per pair (newer listings have shorter history).
- The HF **dataset viewer is disabled** (custom per-pair layout); load files directly.
- Data provided **as-is** for research; no warranty of accuracy or completeness.