Hi I am new to the ML4Trading site. I have been reading some of the blog posts. I am weighing the effort vs the cost of developing a backtesting/simulation capability in Python using the available libraries. The old quantopian libraries appear to be kept up to date on a piecemeal basis and appear to very finicky with regards to the environment and configuration they run in. I am also new to Python.
Q. How stable are the latest versions of Zipline any Pyfolio?
Q. How dependent are the Zipline and Plfolio libraries on the quantopian data bundle format? I would like to simply download data from yahoo and run the code against it.
Zipline and Pyfolio (reloaded) are fairly stable in my opinion.
In my very biased opinion and experience, it’s not the quantopian libraries that are finicky but Windows and MacOS , they usually get in the way of a smooth installation in some way or another .
These days I would even recommend simply using WSL2 for Windows users. I also usually don’t recommend using conda it can be nice for beginners, but sometimes it simply adding another problem.
Having said that, Zipline is a fairly complex piece of code that requires some learning effort.
You can feed zipline with any data you want provided it comes in the right format /calendar.
Cheers,