The argument that Python is "second best in more things" seems like a poor one to me, because it doesn't account for the size of the gap between first and second in some cases.
For one anecdote - during my PhD studies, I knew Python very well, and spent a week or two working with and around SciPy's not-quite-complete stats libraries. Eventually I threw in the towel, used Python as a preprocessor to create CSV files and got it all done in R in a couple of hours of REPL time, with no prior exposure to R.
I'm sure there are thousands of counter examples in which Python stomps all over R for stats analysis, but for my particular use case and set of data, the difference between the best tool and the second best was astonishing.
For one anecdote - during my PhD studies, I knew Python very well, and spent a week or two working with and around SciPy's not-quite-complete stats libraries. Eventually I threw in the towel, used Python as a preprocessor to create CSV files and got it all done in R in a couple of hours of REPL time, with no prior exposure to R.
I'm sure there are thousands of counter examples in which Python stomps all over R for stats analysis, but for my particular use case and set of data, the difference between the best tool and the second best was astonishing.