/cbpro-trader

Automated cryptocurrency trading on Coinbase Pro (formerly gdax-trader)

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

cbpro-trader

cbpro-trader is a bot to automate trading on the Coinbase Pro cryptocurrency exchange. It is based around the ta-lib and coinbasepro-python libraries, allowing easy trading based on technical analysis indicators.

The bot can monitor and trade any ticker supported by Coinbase Pro, and will even trade between cryptocurrencies if the opportunity is better.

Disclaimer

This bot is still in early stages and may have bugs. The strategies also need major reworking. Do not use with any money you are not 100% willing to lose.

Installation

Docker

A docker-compose file is included to try to ease installation.

If you are using the web frontend, you can simply run docker-compose up and then visit the URL provided by the frontend server.

If you are using the curses or debug frontends, you can run

docker-compose build cbpro docker run -it cbpro-trader_cbpro -v $ABSOLUTE_PATH_TO_PROJECT/cbpro-trader/cbpro-trader/daemon/:/cbpro-trader/:Z

to build and run the image, making sure to replace the absolute path of the project, thenpython3 ./cbpro-trader.py to start the bot

You will need to run docker-compose up --build if you change the config after the initial build.

Manual

You can install most requirements with

cd ./cbpro-trader pip install -r ./requirements.txt

then run with

python3 ./cbpro-trader.py

Note that it is currently pointing to a custom version of the coinbasepro-python library until I can push my OrderBook changes upstream.

Also note that the TA-Lib python library is actually a wrapper for the ta-lib C library, so you will need to install that before. See the "Dependencies" section on https://github.com/mrjbq7/ta-lib for more information on that.

Configuration

Copy config.yml.sample to config.yml and include your key, secret, and passphrase values from your Coinbase Pro API key.

Name Type Description
key string Coinbase API key
secret string Coinbase API secret
passphrase string Coinbase API passphrase
sandbox boolean Set to 'yes' to use Coinbase sandbox servers (for testing)
live boolean Set to 'yes' to to toggle live trading
frontend string Which frontend to use - 'console', 'web' or 'debug'. See below for more info.
web_config boolean Set to 'yes' to allow setting config from the web API
logging boolean Set to 'yes' to add additional logging to debug.log file
fiat string Which fiat currency to use - e.g. 'USD' or 'EUR'
max_slippage float Max percentage change in limit orders before executing a market order
periods list A YAML list of periods, each including the options listed in the periods section

For each period in the periods list, include the following options

Name Type Description
name string The display name of the period
product string The Coinbase product to trade
length integer The period length in minutes
trade boolean Set to 'yes' to trade this period
meta boolean Set to 'yes' if this is a 'meta' product (one that does not exist on Coinbase)

Set live to yes only if you want the bot to execute actual trades. The bot will still collect data and calculate indicators when live is set to no. You may also wish to test functionality with a Coinbase Pro sandbox API key first.

In config.yml you can list as many periods as you would like under the periods section. Periods will be used for trading logic only if their trade: attribute is set to yes, otherwise they are just used for gathering indicator data.

There is experimental support for 'meta' periods, which can be used for comparing 2 products that do not currently have a Coinbase Pro trading pair, by setting the meta: attribute to yes in the period description. The only real use case for this right now is LTC-ETH. Trading on meta periods is not yet supported (work in progress).

frontend can have the following values

  • web - a web based frontend to be viewed in your web browser (WORK IN PROGRESS)
  • curses - an ncurses display of balances, indicator values, recent candlesticks and trades and current open orders or debug which will print the same infromation to the console, line-by-line, as it is available.
  • debug used for debugging (obviously).

Tweaking indicators and trade logic

I'm currently working on making indicators and trade strategies more configurable, but if you're handy with Python, any indicators from TA-Lib can be added, as desired. Trade logic can also obviously be modified as well.

Design

The IndicatorSubsystem class contains several methods to calculate indicators. Simple ones are already included, such as calculate_macd() and calculate_obv().

These methods write to the dictionary IndicatorSubsystem.current_indicators which is eventually used by TradeEngine.determine_trades() to determine if the bot should trade.

Adding indicators

To add a new indicator, first create the new method, following the naming convention. For example, if adding Simple Moving Average (SMA), calculate_sma()

In the new method, you will probably want to use TA-LIB to calculate the indicator. Refer to http://mrjbq7.github.io/ta-lib/ for API documentation. For our example:

sma = talib.SMA(closing_prices, timeperiod=10)

Note that closing_prices and volumes are already available in the IndicatorSubsystem.

Now, just add the most recent of this calculated value to the current_indicators dictionary, so that it is available to TradeEngine. You will need to add the indicator to the correct key, determined by period_name as well. This is to support multiple periods.

self.current_indicators[period_name]['sma'] = sma[-1]

Modifying trade logic

Trade logic can obviously be modified as desired. Just make your decisions in TradeEngine.determine_trades().

TradeEngine.determine_trades() has access to the IndicatorSubsystem.current_indicators as indicators, as was discussed earlier.

When issuing a buy order, be sure to set product.buy_flag = True and product.sell_flag = False before starting the buy order thread. This is to be sure that if there is a sell order pending, it will be cancelled and the sell thread will be closed. The same obviously holds true when issueing a sell order.