Stuck while running pair trading strategy backtest

Hi Michael, 

I'll keep this short: 

• Pandas and numpy are great libraries for research, but less suitable for backtesting. Use them as little as possible and if so only where you process larger amounts of data less frequently in batch. 
• Reduce history requests to a minimum. You really only need them for warming up.
• Design your computations as online/one-pass algorithms. Use the indicators provided by QC where possible. You can also modify them with the indicator extensions in many ways to more complex indicators. A concrete example: you calculate the mean every single minute via np.mean(data). This is inefficient. It would be better if you have a rolling sum where you just add the new datapoint and remove the oldest one if length > period. Or even simpler: just use the SMA indicator that QC already provides. This is about 100x faster than calculating the mean again every minute. 
• Refactor your code. It has become very confusing. For instance, many code blocks can be extracted into methods. This won't make your algorithm run faster, but will make the work easier. 
• In general, I recommend to start with smaller and simpler projects and gradually add more elements to your algorithm. This way, weak points will be noticed earlier. 

Hope this helps. 

Hi Arthur, 

That's really helpful. I'll take the notes from your comments and move on. Appreciate!