I've been learning the basics of Python over at DataCamp in preparation for heading over here to Quantopian.
As an early project I downloaded some historical data and wrote some python code to calculate the RSI for said data.
When I ran the code and checked it against the same calculations in Excel, however, I noticed that Excel and Python were returning quite different final results. I traced the issue back to what I believe to be float-point rounding differences/errors in the moving averages stage, that continue to get compounded and end up making a fairly large difference in the final RSI calculation (differences as large as 10 or more in some cases).
I had similar results with MACD calculations.
My question then is, how do you deal with rounding point issues, and how do you know when your numbers are correct? Having numbers slightly off wouldn't be a major issue, but being 10 points off on an RSI calculation can be quite a problem.