CyTrade is a repo with some examples for stream-learning-based Machine Learning strategies for trading cryptocurrency. These examples are built with Cytrader (a cythonized fork of Backtrader), Ct-CCXT-Store (a fork of Backtrader-CCXT-Store) and River.

This repository is accessible at: CyTrade

pip install git+https://github.com/Saran33/_CyTrade.git

• To build from cython files:

cd cytrade
python setup.py build_ext --inplace

Cytrade is set up to use environment variables for API keys, as well as to indicate whether you are live trading or usng a demo account, and to indicate whether you are trading spot or futures. Set a .env file in the cytrade directory, or alternatively set environemnt varibales with your cloud provider. e.g.:

LIVE=False
FUTS=False

BINANCE_KEY=YOUR_LIVE_KEY
BINANCE_SECRET=YOUR_LIVE_SECRET

BINANCE_TEST_KEY=YOUR_TEST_KEY
BINANCE_TEST_SECRET=YOUR_TEST_SECRET

BINANCE_FUTS_TEST_KEY=YOUR_FUTURES_DEMO_KEY
BINANCE_FUTS_TEST_SECRET=YOUR_FUTURES_DEMO_SECRET

• The backtests/btRiver.ipynb notebook file contains an example logistic regression strategy, integrating River with Backtrader.
• It is using the Kelly Critereon to size positions (solving for the maximum growth rate is done in cython, which is 200 times faster than if done in scipy and numpy cytrade/cyutils/cyzers.pyx).
• In the example, the Kelly allocation fraction is scaled by the model's probability estimate.