Implement an actual shuffle function

[RoninTheOrigina]

Yes, you read that right.  The "shuffle" function in Spotify is nothing more than a randomizer.  There's a difference?  YES.

RANDOM - Play one song.  The next song can be ANY song in your playlist.  Including the one you just listened to.  Though I've never had this specifically happen, I literally just heard the same song that I heard 2 songs ago.  This is how Spotify's "shuffle" currently functions.

SHUFFLE - Take your entire list, shuffle them (like cards), and play the first song in the shuffle.  Once finished, move that song off the pile.  Once the pile is depleted, shuffle the deck again and start over (maybe pop a message up).

I know enough about programming (not much) to know that shuffling is probably a bit harder than random, but come on.  I may be in the minority, but I don't have a lot of little playlists.  I have one list with nearly a thousand songs in it, because I like all the music, and am rarely not in the mood for one of those songs.  But what I can't stand is hearing the same song over and over again.  I think there may even be songs in my list I haven't heard yet, while I've heard others 4-5 times.

I know it may not matter much, but I would probably lay down the money for at least a year subscription if this feature was implemented.

EDIT: Another user pointed out that I can see what all is queued up in my shuffle by clicking the Play Queue link, and that Spotify supposedly sets up a rolling shuffle of 50 songs.  It would be nice if this range could be expanded to say 50-100% of your playlist.

EDIT 4/6/12: I just dealt with the same song 16 tracks later.  The song played the first time yesterday afternoon.  I went home, listened for a bit on my home PC (just a few songs), then started listening again this morning.  If the list recycles itself after a day, or if you log in from another computer, I see that as an area for improvement.

[Ericm1030]

They need to make the shuffle less random to be more random, add parameters to make sure whem I scip a song and a new one comes and I don't like it and I skip it the first song come on again.

[CoCavan11]

Three years have passed since the initial suggestion that the Spotify "Shuffle" feature, which was actually a "random" feature, be improved and, from what I can tell, no truly effective improvements have been made.  Essentially, Spotify's "Shuffle" feature still sucks, even moreso in that it's been sucking for at least three years.  I won't even think about upgrading until all the bugs are worked out.  I'll willing to pay for a working system, but I'm not willing to finance a system fix so that I can pay for it.  That's just absurd.

[sgreene822]

I just drove home (a 20 minute trip) and constantly skipped songs that I've heard dozens of times during the few past days. I didn't get to listen to 1 song the entire drive! I'm livid **bleep**ing pissed that I pay for a subscription and this **bleep** is still NOT fixed. Anyone have any suggestions of another streaming music app that is actually worth paying for?

[CoCavan11]

Pandora does have a fairly good system, but I find it too expensive to justify; however, the commercials on the free version aren't sooooooo annoying, given that it's free.  As for just listening to classical music---which is my preference 85% of the time, your best bet is KUSC.  I don't mind donating monthly to this very, very high quality public radio station, particularly since most public radio stations have deteriorated terribly as a result of our nation's poor overall education.  None of the on-air journalists, speakers, announcers, etc., on every NPR station I've listened to in the past five years knows how to pronounce 15% of the words they need to pronounce in order to get through the day.  Not complex, technical, obscure, pretentious words, but common words that good students at good high schools not only could pronounce fifty years ago, but also used in their everyday conversation and in their written school work.  More and more, we are becoming a nation of functional illiterates who can txt NEthng, but understand nothing.

[andrewnwilliams]

The problem is it seems Spotify is shuffling the artists, not the songs. My artists are shuffled, but when an artist is chosen by the shuffler, I usually end up hearing the same songs by that artist. I have huge playlists, and I have songs that I’ve listened/skipped hundreds of times. I’m not sure if this is due to licensing fees or what, but it’s a real problem, and I’ll be looking closely at the Apple’s new Beats service when it launches.

[Dithering_Idiot]

Dear Spotify people,

What your users want is not for songs by the same artist to be evenly distributed, but for the songs themselves to be (somewhat) evenly distributed. Any sort of random song-selecting function (even if the artists are well distributed) will still have short-term patterns in song choice, which for the listener are obnoxious. If we want a certain subset of songs to show up more often, we would like to choose which songs those are (either through duplicates or a smaller playlist or whatever) rather than having a random generator do so. Likewise your random song selector will occasionally ignore a song for much longer than a user would like just by chance. If a user put a song in their playlist, that means they'll want to listen to it about as often as any other song, rather than, by chance, not hearing it at all for a long interval. The idea of a real shufffle function is a card deck. The songs are shuffled randomly, played through, shuffled again, etc. While there is a small chance that the same song may be played twice in a row, it guarantees that, over any listening time interval, no song can be more than 2 plays ahead of or behind any other song (which is more evenly distributed, and thus more pleasing to the user, than your current algorithm). And you can guarantee that no song ever shows up twice without playing through at least a 1/n fraction of all the songs in the playlist pretty easily. Take the deck of songs that you've just played through. Remove the 1/n fraction that was just played. Now shuffle the remaining songs, and set aside the top 1/n fraction. Shuffle in the 1/n fraction that was just played and add on top the 1/n songs you set aside. And there you go. Evenly distributed but still somewhat random song choice with a good distribution of distance between each song play. You could even keep your fancy artist shuffling algorithm if you want.

[user-removed]

If Spotify is not going to resolve this issue I'm going to stop my premium subscription no use in Spotify playing the same songs I'd rather listen to the darn radio

[mlidbo]

This is driving me mad. 

I have a large collection of songs.

How large? I ran into the 10000 song limit and had to delete a lot of saved songs.

That large....

I almost always just go to songs and click "shuffle". 

So why do I hear the same tired songs again and again!? (Cheaper royalties or more weight to locally cached songs, and the paying customers can go f  themselves?)

It's like spotify refuses to ever play 90% of my saved songs, and plays a few songs so often that I get royally sick of them.

I'm am very seriously considering leaving spotify for another service. What does it matter that spotify in principle has more of the songs I like if I never get to hear them!?

Like so many basic Must Work features spotify apparently continues to ignore that it is utterly broken and keep on building and tearing douwn huge useless creations such as the app store or whatever they called it. I never used it.

I do listen to my collection "shuffled" for hours every day (How about actually playing it shuffled instead of a few songs?)

I do add and remove songs every day (How about not silently breaking completely without warning if I happen to get more than 10000 songs?)

How about fixing the playback on android so that it does not stop/stutter all the time even though my connection tests as having more than 30Mb bandwidth?

How about fixing how the playback occationally just goes silent with the seconds ticking up?

How about releasing an updated version on the linux client that has spotify connect?

Oh, one positive. Major cudos for fixing spotify connect so that it works as it should. You bring your active playlist along with you as you switch devices. Excellent. I use it all the time now while I never used it before 🙂

Not sure why I post this. If anyone from spotify ever actually read comments here this would be fixed already. Maybe they occationally read the first few comments on new ideas/issues. But there is just about 0% chance anyone from spotify will read this. 

[mlidbo]

Proof that this is broken

If you will grant me that I am not lying I can prove that it is broken. (Feel free to confirm this using data you have about my account: magnus_lidbom).

Go to my last.fm profile here: http://www.last.fm/user/EsotericMange

Click "last 7 days"

The top plays/artists are:

Plays Artist

24 Guns n' roses.

21 Jamiroquai

21 India Arie

16 Saga

11 Pearl Jam

...

I have not played any playlist with any of those artists on it in the last 7 days!

I just play "Songs" or sometimes Favorites using "Shuffle" So why are those artists so bizarrely overrepresented? The number of songs by those artists that I have in Songs and Favorites in no way corresponds to this percentage of plays. Nowhere near. This is an obvious skewing of the selection of songs.

[mlidbo]

If you want the data in a computer readable format so that you can analyze it easily using code there is an RSS feed here:

http://ws.audioscrobbler.com/2.0/user/EsotericMange/weeklyartistchart.xml

Strangely the numbers are different here. I'm guessing the data on the page is is a snapshot from a different point in time than the data in the feed. The same thing goes though. Some artists are hugely overrepresented for no sane reason: 

20 Jennifer Lopez

19 Saga

17 Bo kaspers orkester

15 Dido

15 Morcheeba

15 Pearl Jam

Again: I have not played any playlist with any of those artists on it in the last 7 days!

The only possible source for them is Songs or Favorites, where they are in no way represented in percentages similar to the persentage of plays logged by last.fm. 

EDIT: Here is a link to the same statistics as in the feed (as it looks right now. Apparently a snapshot from the week ending on 2015-02-29) using a link where the data should not change:

http://www.last.fm/user/EsotericMange/charts?charttype=weekly&subtype=artist&range=1427025600-142763...

Unfortunately that is not machine readable.