Algotrading Framework is a repository with tools to build and run working trading bots, backtest strategies, assist on trading, define simple stop losses and trailing stop losses, etc. This framework work with data directly from Crypto exchanges API, from a DB or CSV files. Can be used for data-driven and event-driven systems. Made exclusively for crypto markets for now and written in Python.

A Medium story dedicated to this project

• Operating modes

  • Realtime
  • Tick-by-tick
  • Backtest

• How to start

  • Pre Requesites

  • Collecting data

    • Database
    • Script

• Entry functions

  • cross SMA

• Exit functions

  • cross SMA

• Plot data

• Logs

• Examples

  • Realtime alert for volume increasing
  • Backtest a strategy on BTC-DGB pair using SMA

• Pypy

• TODO

• Additional info

Framework has three operating modes:

• Realtime -- Trades with real data in real time, with real money or in simulation mode.
• Tick-by-tick -- Testing strategies in real time frames, so user can follow its entries and exits strategies.
• Backtest -- Backtesting strategies and presenting the results.

In realtime, Trading Bot operates in real-time, with live data from exchanges APIs. It doesn't need pre-stored data or DB to work. In this mode, a bot can trade real money, simulate or alert the user when its time to buy or sell, based on entry and exit strategies defined by the user. Can also simulate users strategies and present the results in real-time.

Tick-by-tick mode allows users to check strategies in a visible timeframe, to better check entries and exit points or to detect strategies faults or new entry and exit points. Use data from CSV files or DB.

Allows users to backtest strategies, with previously stored data. Can also plot trading data showing entry and exit points for implemented strategies.

To get algotrading fully working is necessary to install some packages and Python libs, as IPython, Pandas, Matplotlib, Numpy, Python-Influxdb and Python-tk. On Linux machines these packages could be installed with:

pip install -r requirements.txt

The first step is to collect data. To get markets data is necessary to run a DB, to get and manage all data or save the data to CSV files. There are two options:

• Install and configure a database.
• Run a script to collect data and save it to CSV files.

Trading Bot is ready to operate with InfluxDB, but can work with other databases, with some small changes.

To install, configure and use a InfluxDB database, you can clone this repository: https://github.com/ivopetiz/crypto-database

If you don't want to install and manage any databases and simply want to get data to CSV files you can use the script in this Gist: https://gist.github.com/ivopetiz/051eb8dcef769e655254df21a093831a

Using a database is the best option, once you can analyse and plot data using DB tools, as Chronograf, and can always extract data to CSV if needed.

Entry functions aggregate all strategies to enter in a specific market. Once data fill all the requisites to enter a specific market, an action is taken. Users can use one or several functions in the same call, to fill the requisites and enter market/markets. Functions should return True, if the available data represent an entry point for the user. If not, the return needs to be False. <entry.py> should aggregate all users entry functions.

Function <cross_smas> will return True if first SMA cross the second one. If not will return False.

def cross_smas(data, smas=[5, 10]):
    '''
    Checks if it's an entry point based on crossed smas.
    '''
    if data.Last.rolling(smas[0]).mean().iloc[-1] > \
       data.Last.rolling(smas[1]).mean().iloc[-1] and \
       data.Last.rolling(smas[0]).mean().iloc[-2] < \
       data.Last.rolling(smas[1]).mean().iloc[-2]:
        return True

    return False

Exit functions have all functions responsible for exit strategies. When a user is in the market, and data met exit criteria, the bot will exit the market. Exit functions can be used with other exit functions, to cover more situations, as used in entry functions. Stop loss and trailing stop loss are also implemented, to exit markets in case of an unexpected price drop. Functions should return True, if the available data represent an exit point for the user. If not, the return needs to be False. <exit.py> should aggregate all users' exit functions.

Function <cross_smas> will return True if first SMA cross the second one. If not will return False.

def cross_smas(data, smas=[10, 20]):
    '''
    Checks if it's an exit point based on crossed smas.
    '''
    if data.Last.rolling(smas[0]).mean().iloc[-1] < \
       data.Last.rolling(smas[1]).mean().iloc[-1] and \
       data.Last.rolling(smas[0]).mean().iloc[-2] > \
       data.Last.rolling(smas[1]).mean().iloc[-2]:
        return True

    return False

It's possible to plot entry and exit points, among market data, using Matplotlib lib for Python with the option plot=True on function call.

Can log entry and exit points in order to evaluate strategies, presenting P&L for specific markets and total.

Here are some examples of how to use this framework.

To get an alert when a market doubles its volume:

from cryptoalgotrading.cryptoalgotrading import realtime

def alert_volume_x2(data):
    if pd.vol.iloc[-1] > pd.vol.iloc[-2]*2:
        return True
    return False

realtime([], alert_volume_x2, interval='10m')

alert_volume_x2 checks the value of actual market volume and compare it with the last time frame volume value, alerting user when actual market volume is bigger than last time frame volume value multiplied by 2. Can add functions live on IPython for example of add them to entry and exit python files.

To backtest a cross simple moving average strategy in a specific market and plot the entry points:

from cryptoalgotrading.cryptoalgotrading import backtest
import cryptoalgotrading.entry as entry

backtest(["BTC-XRP"], entry.cross_smas, smas=[15,40], interval='10m', from_file=True, plot=True)

Based on market data available for BTC_XRP pair, code above can present an output like this:

The figure has three charts. The chart on top presents on top BTC-XRP data from a certain period, with its Bollinger bands and 3 SMA lines. Green points represent the entry points for the defined strategy. In the middle is a chart representing volume data and at the bottom is represented the number of selling orders among time. All these fields and charts are configurable on plot function.

Can also add exit points by adding an exit function or functions to backtest function. It is possible to enter multiple entries and exit functions to backtest, to define different entry and exit positions.

Both functions are available on entry.py and exit.py as example.

In finance.py are some functions which could be useful to implement some strategies.

This Crypto AlgoTrading Framework can be used with Pypy, but the results will not be great, during the use of Pandas and Numpy libs.

• Add ByBit API
• Python 3 implementation
• Cython version
• Pure Python for Pypy (without Pandas and Numpy dependency)

API Key is just needed in case of buy/sell operations. For backtest, tick-by-tick and realtime alert implementations API Key can be left empty.

Buy and sell options are commented and should only be used if you know what you are doing.

USE THIS AT YOUR OWN RISK.