Download the datasets (ZIP file, 8MB).

For a while, I had been building a neural network that could predict the arbitrage opportunities in various Bitcoin exchanges. Unfortunately I have no time currently for this project, so I have now decided to publish my datasets I had used for this project.

In the ZIP file, there are four JSON files for 4 major Bitcoin exchanges (GDAX, Kraken, Bitstamp & Bitfinex). The data is an array of length of 1752147 indexes, 0 to 1752146. Each value is a minute interval, starting from 1st of December 2014, ending somewhere in early 2018. The values are the current "ask" prices in each exchange.

The indexes are matched across the datasets, so for example value at index 10000 in each file, is the same time in all exchanges. This allows you to see the differences between exchanges and analyze the arbitrage opportunities.

Hope you have fun with these datasets, and please let me know if you use them in something interesting!

Example of how to read them in Python:

import numpy
import json

# load the dataset
dataframe_gdax = numpy.array(json.load(open("datasets/gdax-usd.json")))
dataframe_kraken = numpy.array(json.load(open("datasets/kraken-usd.json")))
dataframe_bitstamp = numpy.array(json.load(open("datasets/bitstamp-usd.json")))
dataframe_bitfinex = numpy.array(json.load(open("datasets/bitfinex-usd.json")))

Source of datasets: Bitcoincharts.com.