Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Spotify's awful discovery, and shuffle features for playlists are what keeps me firmly with Pandora.

Now that Pandora lets you make your own playlists, or listen to individual songs, and download for offline playback (on airplanes for example), I don't see any compelling reason to switch to Spotify these days.



Anecdotal, but Spotify's Discover Weekly feature has ridiculously good recommendations for me and I usually end up adding half of each week's playlist to my library. The recommended music is fantastic, obscure stuff I would never have known about otherwise. My experience with Spotify recommendations couldn't be more different to yours.


I'll absolutely agree with this. The Discover Weekly playlist seems to know my musical preferences better than I do myself, and i'm certainly not sticking to just a couple of specific genres that I listen to


I enjoyed Discover Weekly at first but over time got sick of only hearing suggestions from inside my own “bubble”. I’ve switched to human curated radio and the variety is so much better. haven’t looked back.


Fair enough. Have you tried Pandora's for an extended period and are able to compare?

For me, Spotify ended up just playing the same artist, same album even, over and over. Same problem with their Shuffle feature - I could have a playlist with 500 songs from various artists, albums, genres and more... but I'd get the same artist, same album back to back to back far too often. So often, it didn't feel random at all. Just anecdotal though.


They are random! It's just that what is random in a mathematical sense isn't random to a human.

https://en.wikipedia.org/wiki/Clustering_illusion

https://en.wikipedia.org/wiki/Poisson_clumping

The best song randomizer I've seen described generates a permutation of songs, then "plots" them in n dimensional space with closeness being closeness being a weighted distance of closeness in the playlist and whatever metadata you have about them. (artist, genre, year, mood). Then you iteratively break the clumps until you can't find any or the number stops decreasing. If you still have clumps then you raise the bar for what is considered a clump and repeat until you hit zero.


And they wrote a blog about it in 2014: https://engineering.atspotify.com/2014/02/28/how-to-shuffle-...

Interesting stuff!


That is an interested blog post!

However, I've used Spotify since 2014 and still experienced the same Shuffle behavior. Strange, perhaps I'm looking for it now, and introducing some sort of bias.

It is interesting, however, that a music playlist company took so long to realize people wouldn't like their original Shuffle implementation. I don't think any device or software I used before Spotify had such a remarkably bad Shuffle experience - including the original iPods and other, much older software and devices.


True. It's just that I don't think I've ever experienced Spotify's take on Shuffle anywhere else before, so it's shocking and frustrating, leading to a lot of "next, next, next" skipping.


I think that Pandora's recommendation results are far superior to Spotify's, thanks to the Music Genome Project. I think Spotify does some basic "A likes X and B likes X, so if A likes Y, B will like Y" processing. My ideal-new music-discovery workflow gets the best of both worlds: find stuff on Pandora, and use the results to create playlists on Spotify.


Last.fm's recommender system continues to be the only worthwhile music recommendation service.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: