I've always been interested in stocks and trading but I had never attempted any sort of project in it. So after some reading about different concepts and picking a route I wanted to go down I decided to commit to building a trading platform. What I mean by a trading platform is a system that includes a strategy and a way of backtesting this strategy. And well, I've finally got a working a platform. It's far from perfect, this is only really the beginning but it has a functioning strategy and a backtester.

What is the strategy used?

I discovered the concept of pair trading and cointegration by accident. But it instantly struck me as an interesting introductory project for me as it had a nice amount of statistics in it. I've decided to start off with the most basic form of a cointegration pairs trading strategy I could think of check if a pair is cointegrated over the last n months, if it is then over the next k months, long the spread when it is low, and short the spread when it is high, closing positions as it approaches the mean.

Code layout:

I’ve broken the code into 4 different sections:

• Getting the data (DataManager.py)
  
• Analysing the data for cointegration (cointAnalysis.py)
  
• Creating the strategy (cointStrategy.py)
  
• And finally, backtesting (backtester.py).
  

Future Plan:

• Implement a stop loss
  
• Use rolling windows instead of fixed
  
• Baskets instead of pairs
  
• Look for cointegration over several windows not just one
  
• Hurst exponent (or Half Life)
  

For more details about a similar strategy see this video: https://www.quantopian.com/lectures/introduction-to-pairs-trading