https://www.apple.com/newsroom/2005/09/07Apple-Introduces-iT...
My prime conjecture now is that there's some kind of caching reason where it's more advantageous for their CDN if those handful of songs are the only ones that're played. Funnily this also happens in my offline playlist, but I guess this is just because the same algorithm is also used there.
Shuffling music on a per song basis is an interesting problem, with the ideal solution being something a top class DJ would do, e.g. matching one characteristic of the two tracks (e.g. Tempo), but updating another (e.g. Timbre).
But even with that implemented, not everybody likes the same thing. E.g. I love to hear new music that I don't know yet, but as a DJ my experience has been that many people like to hear music they know. Meaning app developers have to make the right choices available and those choices are under the hood far more complex than how you can sort a list of tracks by artists.
Exactly. There is no winning here.
When I was in college, I built a system for the college radio station that plays music while there are no DJs on the air. What I enjoyed most is keeping track of what songs played when (which we needed to do anyway for FCC compliance) so that we never play the same song twice until every song was played. This "felt" more random even though it wasn't.
Some people didn't like it. They wanted to hear a smaller subset of songs more often (the equivalent of putting a playlist on random). We solved that by letting users request songs through the website.
I wonder if there's a way to solve this by adding a setting called "repetitiveness". It's a value between 1 and 100. 1 is the least repetitive "play all songs once before repeating any" and 100 is "play the same sequence of songs every time."
https://web.archive.org/web/20230410041435/https://engineeri...
EDIT:
> We noticed some users complaining about our shuffling algorithm playing a few songs from the same artist right after each other.
...
> Since the Spotify service launched, we used Fisher-Yates shuffle to generate a perfectly random shuffling of a playlist.
...
> The algorithm is now rolled out to everyone using our desktop client and other clients will follow soon.
Everything old is new again.
That's the one! Thank you!
https://www.music-map.com/info
From running the music discovery websites Gnoosic and the Music-Map for years now, and doing a lot of experiments, my experience is that the chance of someone liking a truly randomly selected band is very low. Less than 10%. But hey, trying 20 random bands and finding one interesting one can be worth it.
It's far more sinister than this. It has to do with royalties. They've got some secret algorithm that will even cut your account off from specific content if it's expensive and you consume it too frequently.
I find no issues listening to pop cult shit 24/7/365, but when I want to listen to some obscure opera more than 3 times it inevitably starts to fade out like a ghost.