Bitcoin Arbitrage Trading Bot

A python monitoring and trading bot which exploits price-spreads between different cryptocurrency exchanges.

Capabilities

• Monitoring of spreads between exchanges (eg. Gdax, Bitfinex, Bitstamp, ...)
• Multiple currency pairs configurable (eg. BTC/EUR, ETH/USD, ...)
• Notifications on spread thresholds (eg. spreads of > 200€) via different notification channels (eg. Pushbullet, Mail, ...)
• Automated trading for configured spread thresholds (partly implemented)
• Storing of historical spreads and making them available via a web UI
• Highly configurable (currency pairs, thresholds for each notification channel or trading, historical data, ...)

Trading Strategy

The following example explains spreads between exchanges:

This results in a spread of 600€ for the price of 1 BTC in EUR between Gdax and Bitfinex (this example is not based on ask and bid prices for simplicity reasons, the bot itself uses those correct prices).

How to exploit this spread?

(example based on above table)

Strategy 1 (common "arbitrage trading")

Preparation:

• Store 5000€ on Gdax

Spread exploitation:

• Buy BTC on Gdax for 5000€
• Send BTC from Gdax to Bitfinex
• Pay fees for transaction and wait until it's confirmed
• Sell BTC on Bitfinex for 5600€

Postprocessing (steps needed to be able to repeat exploit):

• Transfer 5600€ from Bitfinex to Gdax to repeat

Strategy 2 (used by this trading bot)

Preparation:

• Store 5000€ on Gdax account
• Store 1 BTC on Bitfinex account

Spread exploitation:

• Buy 1 BTC on Gdax
• Simultaneously sell 1 BTC on Bitfinex

Postprocessing (steps needed to be able to repeat exploit):

• Transfer 5600€ from Bitfinex to Gdax
• Transfer 1 BTC from Gdax to Bitfinex

The following table demonstrates the amount of BTC and fiat stored on the exchanges during the different phases of the execution of the strategy:

Preparation

After Spread exploitation

Profit: 600€ (excluding fees for simplicity)

Status of project

The strategy worked during the crypto boom at the end of 2017 (with spreads of > 700€ several times a week). As of now, spreads are pretty low (< 100€) which renders using the bot risky and less attractive (change of prices during order executions, fees, ...).

• Monitoring works
• Automated trading is partly implemented, but never tested with real accounts
• Reliability isn't as good as it should be when using the bot with real accounts

Developer Information

The bot itself is placed at bitcoin_arbitrage/monitor with its entry point / main class being bitcoin_arbitrage/monitor/monitor.py.

The code within bitcoin_arbitrage/ is a Flask app which is just a fancy wrapper for the monitor/bot with a web UI.

Configuration of the bot is done by copying the settings.py file.

Running locally

Requirements:

• python >= 3.6
• pipenv (install via pip install pipenv)

Run the monitor:

• pipenv install - Install requirements
• scripts/copy-config - Copy config file (some dummy settings are set by default)
• scripts/run - Run the monitor

Running on server

You can use the following systemd service entry:

[Unit]
Description=Bitcoin Arbitrage Monitor
After=network-online.target

[Service]
Type=simple
User=bitcoin
Restart=always
WorkingDirectory=/home/bitcoin/bitcoin-arbitrage
ExecStart=/home/bitcoin/bitcoin-arbitrage/scripts/run

[Install]
WantedBy=multi-user.target

Running the tests

Manually:

• scripts/test

With PyCharm: