Kraken Market Data

Real-time and historical market data from Kraken exchange, optimized for ML/AI trading research.

🆕 Latest Update (June 2026)

• Added SPCXx/USD (SpaceX) - Tokenized SpaceX stock, newly listed on Kraken
• Real-time OHLC data captured via WebSocket from listing date

Previous Updates

January 2025

• Added PAXG (PAX Gold) - Tokenized gold data for Gold vs Bitcoin analysis
• New pairs: PAXG/BTC, PAXG/USD, PAXG/EUR, PAXG/ETH
• Historical OHLC data across multiple timeframes

Dataset Structure

data/
├── crypto/
│   ├── ticker/     # Real-time price snapshots
│   ├── book/       # Order book depth (100 levels)
│   ├── ohlc/       # OHLC candlestick data
│   └── trade/      # Individual trade executions
├── ohlc_1m/        # 1-minute candles
├── ohlc_5m/        # 5-minute candles
├── ohlc_15m/       # 15-minute candles
├── ohlc_1h/        # 1-hour candles
├── ohlc_4h/        # 4-hour candles
└── ohlc_1d/        # Daily candles

Available Pairs

Cryptocurrencies

PAX Gold (Tokenized Gold)

Note: PAXG represents 1 troy ounce of London Good Delivery gold stored in Brink's vaults.

Tokenized Stocks (xStocks)

Note: xStocks are tokenized representations of real-world stocks traded 24/7 on Kraken.

Data Schema

OHLC Data (Parquet)

Ticker Data (Parquet)

Order Book Data (Parquet)

Usage Examples

Load OHLC Data with Pandas

import pandas as pd
from datasets import load_dataset

# Load daily PAXG/BTC data
ds = load_dataset("Abraxasccs/kraken-market-data", data_files="data/ohlc_1d/PAXG_BTC*.parquet")
df = ds['train'].to_pandas()

# Calculate returns
df['returns'] = df['close'].pct_change()
print(df.tail())

Load Multiple Timeframes

import pandas as pd

# Load from HuggingFace
base_url = "hf://datasets/Abraxasccs/kraken-market-data/data"

btc_1d = pd.read_parquet(f"{base_url}/ohlc_1d/BTC_USD_1d_history.parquet")
btc_4h = pd.read_parquet(f"{base_url}/ohlc_4h/BTC_USD_4h_history.parquet")

print(f"Daily candles: {len(btc_1d)}")
print(f"4H candles: {len(btc_4h)}")

Gold vs Bitcoin Analysis

import pandas as pd

base_url = "hf://datasets/Abraxasccs/kraken-market-data/data"

# Load PAXG/BTC ratio
paxg_btc = pd.read_parquet(f"{base_url}/ohlc_1d/PAXG_BTC_1d_history.parquet")
paxg_btc.set_index('timestamp', inplace=True)

# Calculate z-score for mean reversion
mean = paxg_btc['close'].rolling(50).mean()
std = paxg_btc['close'].rolling(50).std()
paxg_btc['zscore'] = (paxg_btc['close'] - mean) / std

# Find extreme readings
overbought = paxg_btc[paxg_btc['zscore'] > 2]
oversold = paxg_btc[paxg_btc['zscore'] < -2]

Data Collection

• Source: Kraken Exchange WebSocket API v2 & REST API
• Update Frequency: Real-time (WebSocket) + Hourly (REST OHLC)
• Storage: Apache Parquet format
• Timezone: UTC

License

MIT License - Free for commercial and non-commercial use.

Citation

@dataset{kraken_market_data_2025,
  author = {Lugo Munoz, German Gregorio},
  title = {Kraken Market Data},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/datasets/Abraxasccs/kraken-market-data}
}

Related Projects

• Kraken API Documentation
• PAX Gold (PAXG)