Hi, author here! Interesting question!

Martin Fiedler briefly addresses this topic in a comment on his blog post about shuffling algorithms (https://keyj.emphy.de/balanced-shuffle/)

> Apple has a so-called “smart shuffle” algorithm, but this merely puts higher-rated tracks in front of lower-rated ones. So basically, it’s just random shuffle, followed by a sort-by-rating operation. I don’t know of any product (software or hardware) that uses some kind of smart, balanced or whatever-you-like-to-call-it shuffle based on the principles I described in the text.

I'm not sure what they are using today.

What kind of categorization did they use to keep it from just becoming a sort? By artist?

I just found this 2009 research paper written by a group of statisticians titled 'Does Your iPod Really Play Favorites?' (https://www.researchgate.net/profile/Jessica-Culhane/publica...).

In this paper, they test different probability models to detect bias in iPod's shuffling algorithm and eventually conclude that:

> Our statistical tests show the long-term occurrences of these events are within expectations under the assumption of a random shuffle."

Regarding sorting by artists or groups, they found that:

> We failed to find any evidence to support the claim of users like Steven Levy of favoritism of certain groups in the shuffle.

No, I mean if you order from most to least popular it's always the same. Did they pick a random artist then sort the artist's tracks by popularity?