An interesting note: Trevor Hastie is an author on this paper. The crowd around here probably knows him best for books he co-wrote: The Elements of Statistical Learning (2001) and An Introduction to Statistical Learning (2013).
It's funny reading the comments in this thread from random internet people postulating that the authors have probably made some statistical mistake or overlooked something. Then you see Hastie in the author list. Nope, no methodological mistakes made here.