edouardpoitras/nowtrade

Python library for backtesting technical/mechanical strategies in the stock and currency markets

NowTrade

NowTrade is an algorithmic trading library with a focus on creating powerful strategies using easily-readable and simple Python code. With the help of NowTrade, full blown stock/currency trading strategies, harnessing the power of machine learning, can be implemented with few lines of code.

NowTrade strategies are not event driven like most other algorithmic trading libraries available. The strategies are implemented in a sequential manner (one line at a time) without worrying about events, callbacks, or object overloading.

The Basics

All strategies follow this basic pattern:

• Load the required data for your strategy
• Train your machine learning algorithm (optional)
• Add your technical indicators (example: moving average, relative strength index, neural network, random forest, etc)
• Add criteria based on your technical indicators (example: neural network prediction is higher than current price)
• Associate actions to those criteria (example: go long when a certain criteria is true)
• Create a trading profile (includes information such as your capital, trading amount rules, trading fee rules, slippage rules, etc)
• Send all of this information to a strategy simulator and wait for results

Check out some example strategies in the examples directory.

Features

Here is a list of reasons you may want to use NowTrade over alternatives:

• Implement strategies easily
• Flexible data connections (store data locally from any remote data connection)
• Easily implement machine learning into strategies (neural networks, random forest, etc)
• Simple, small, and clean code base (build failure on pylint warning)

Cons

List of reasons you may not want to choose NowTrade over alternatives:

• Not mature yet
• Performance sacrificed for simple abstraction and code base
• Still lacking many features (contributions welcome!)

Installation

NowTrade has only been tested on Ubuntu 14.04 and 16.04. It will most likely run just fine on any UNIX-based operating system provided all dependencies have been met.

Dependencies

apt-get install python-pip python-numpy python-pandas python-scipy cython
pip install pandas-datareader

TaLib for Technical Indicators

wget -O ta-lib-0.4.0-src.tar.gz http://sourceforge.net/projects/ta-lib/files/ta-lib/0.4.0/ta-lib-0.4.0-src.tar.gz/download
tar xvzf ta-lib-0.4.0-src.tar.gz
cd ta-lib/
./configure --prefix=/usr
make
make install
pip install Ta-Lib

Matplotlib for pretty charts

apt-get install python-matplotlib

For MySQL Support

apt-get install python-mysqldb

For MongoDB Local Storage Support

apt-get install mongodb python-pymongo

Note: You may need to un-comment the line '#port 27017' in /etc/mongodb.conf for NowTrade to communicate with mongo

For ensemble (random forest) support

apt-get install python-sklearn

For neural network support

pip install pybrain

Finally, install nowtrade

python setup.py install

Tests

To run the tests you'll need nose:

apt-get install python-nose

Then you can run the tests by running the following command in the root of the repository:

sh tests/test_all.sh

Give it a Spin

You can try out some strategies in the examples directory.

I recommend taking a look at the following examples:

• Learn how to pull from remote data sources and storing stock data locally for backtesting purposes
• A simple moving average crossover strategy
• A simple strategy utilizing the random forest machine learning technique
• A simple strategy utilizing the neural network machine learning technique
• Display some pretty charts of your strategies

See Also

You may also be interested in these other python-based backtesting systems:

• zipline - https://github.com/quantopian/zipline
• PyAlgoTrade - http://gbeced.github.io/pyalgotrade/

Author

Edouard Poitras [email protected]

nowtrade's Issues

Need an overview diagram of the entire NowTrade system and it's components

Will make it easier for new comers to contribute and to understand how NowTrade works.
Maybe we should also add flow charts for the strategy creation process.

Replace Position criteria with Above/Below criteria

Position criteria is a bit more convoluted than necessary. Replace with a Above and Below criteria.
Instead of Position(technical_indicator1, 'below', technical_indicator2), we should have Below(technial_indicator1, technical_indicator2).
Same thing for the Above criteria.

Add a Slope criteria to see if a technical indicator or series is moving up/down

Should have a value for how steep the slope should be for the criteria to trigger.

Rename or replace TimeSinceAction criteria

TimeSinceAction is not named properly. It insinuates that we will be able to specify a time (like 10 minutes, or 5 hours) since an action occurred.
It should either be renamed to something more accurate (like BarsSinceAction?) or replaced with more specific criteria such as:
LongOccured/LongHappened/BarsSinceLong
ShortOccured/ShortHappened/BarsSinceShort
LongExitOccured/LongExitHappened/BarsSinceLongExit
ShortExitOccured/ShortExitHappend/BarsSinceShortExit

Add some images to the examples

Some of the examples are pretty bland. We should at least add some chart images in the figures.md example.

Attain 100% Test Coverage

Higher priority than adding features.

Add Hurst Exponent as TI

See https://www.quantstart.com/articles/Basics-of-Statistical-Mean-Reversion-Testing for implementation.

Above/Below/Equals criteria should have a lookback option

To allow criteria such as "Was the SMA10 bigger than the SMA50 5 days ago?".

Better way of managing configuration options

We need a better way of handling configuration options than the configuration.py file in the nowtrade library folder. The user should have the options of having a nowtrade.conf file in their PYTHONPATH that would be parsed every time NowTrade is imported.

There should be sensitive defaults for when the configuration file is not present.

Add a Dockerfile

Should be able to try it directly from a docker run command.

git clone the repo or will make handle that?

Hello,

Just to clarify, the NowTrade repo needs to be cloned to install it right? Or will makefile handle that?

Thanks!

Remove all print statements in code

Should be able to replace them with raise and logger warnings.

pylint should return no errors or warnings

Will eventually remove the --errors-only flag when running pylint. Code should be clean.

Easy way of identifying next action in market

NowTrade is primarily a backtesting system. However, it would be nice to be able to access a 'next_action' from a strategy previously simulated. This would be used for active trading in real markets.

For example, if a user had picked out a strategy on a day time frame, the user could simply re-run the strategy with the latest market data every day and extract the 'next_action' from the strategy results.