News, info, and opinion by Mac users, for Mac users.

March 13, 2007

itunes

Is it randomness or favoritism?

Posted Mar. 13, ’07, 11:39 AM PT by Kate Marshall
Category | iTunes

indexshuffleicon.jpgDoes the shuffle function in iTunes play favorites? It’s not the first time this question’s been asked, even for Newsweek columnist Steven Levy. In theory, if you tell iTunes to shuffle your music library, you shouldn’t hear two Grey Eye Glances sandwiching a Sarah McLachlan tune, followed by three Creed singles (if you’re trying to draw conclusions about me based on this tiny sample of my iTunes library, please don’t. My musical tastes are far more eclectic and dorky than the examples I’ve provided here). Anyhow, CNet Australia’s David Braue set out to answer the question of whether iTunes’s random playback is really that random. His experiment began by adding 200 songs—half purchased from the iTunes store and half ripped from many CDs—to a new Mac mini’s empty iTunes library. Purchased songs came from the four major music labels, with some also showing up on Billboard’s Top 50 chart, in case iTunes was designed to favor pop songs over other genres, or DRM-protected music over regular MP3s. A total of 20 playlists with 650 possible song positions were created using Smart Playlists, and the playback results later analyzed. So what were some of the statistics Braue turned up?

Lionel Richie (Universal) was iTunes’ favourite artist; his songs were chosen 59 times for 40 playlists [iTunes songs only]. Times per possible playlist (TPP) = 1.475. Def Leppard (Universal) was iTunes’ favourite artist among songs ripped from CD; their songs were chosen just 24 times for 20 playlists [iTunes songs and ripped MP3s]. TPP = 1.2
I’m sure there’s a Nicole Richie joke in there somewhere but I’m not going to make it. Either way, it seems that when you ask your Magic 8-Ball about the random nature of iTunes, the best answer you can hope for is, “Repy hazy. Try again later.”


8 Comments

Goose said:

In theory, if you tell iTunes to shuffle your music library, you shouldn’t hear two Grey Eye Glances sandwiching a Sarah McLachlan tune, followed by three Creed singles.

Why not? It could randomly choose to play the songs in that order. Or it could choose to play the songs in the order of Creed x3, Sarah McLachlan and Grey Eye Glances x2. It'd still be random. If it's totally random, combinations like those are perfectly acceptable, since it randomly chooses how to play the songs, and it randomly happens to play it like that.

March 13, 2007 11:56 AM
Ward Author Profile Page said:

You'd think he could at least consult with a statistician... What a horribly flawed analysis. Sample size was small and he made no effor to quantify the significance of his observations.

Occam's razor says that iTunes is random (or at least as random as the psuedo-random number generator allows), unless told not to be. Until somebody does a real study of iTunes' picks, there's no reason to believe that Apple's choosing your music for you.

March 13, 2007 12:11 PM
jayH said:

steve jobs fixed the probability of songs from the same CD being played consecutively with the release of an update for iTunes that allows u to chose how likely that will happen

he made it clear that by making this, u are actually making the shuffle LESS random and u could see in his face how agaisnt using this feature he was haha gotta love him.

March 13, 2007 12:41 PM
Juha Haataja said:

Of all the possible instruments to estimate randomness the human brain is one of the worst - we see patterns even where there are none. Good comment by Ward about estimating randomness - it takes skill to do this right.

March 13, 2007 1:33 PM
Dave said:

Numb3rs had a nice discussion of randomness a while back. He showed two charts with dots laid out with no apparent order. One had clumps the other didn't. For some reason everyone in the class thought the one with clumps was less random. We had a nice demo in a training class here where people were supposed to select random numbers, it's a great way to select primes, not a random sample space.

In short, Goose is right, there is no rule that says your sample play order isn't random. By definition that order is as likely as any other.

March 13, 2007 1:50 PM
Eric said:

One of the misconceptions of randomness is that things will be evenly distributed. In fact, if you had the same relative representation of artists per playlist, I would suspect that things were not random. Or, to put it another way, if 10 people flipping 10 coins each all had exactly 5 heads and 5 tails, I would be suspicious.



When people talk about random play, they don't really want random. They want what they perceive to be random... which is the feature that they added to iTunes awhile back.

March 13, 2007 1:50 PM
Thomas GvL Author Profile Page said:

I think Steven Levy concluded it well in his iPod book ("The Perfect Thing"). Basically, iTunes' randomness is not completely, 100%, militarily random, but it's really good enough. Any patterns we see are really created by our brain. As it was mentioned, random doesn't mean perfectly distributed.

March 13, 2007 3:05 PM
pdkoenig Author Profile Page said:

Actually, inexplicably, they made 20 playlists from 100 purchased songs and then made 20 more playlists from 200 songs including 100 mp3s and the original 100 purchases. That led them to be confused and make a lot of errors -- those TPP numbers are meaningless garbage. Recalculating, it actually appears that they showed that iTunes does, in fact, behave randomly.

March 14, 2007 12:34 AM

Leave a comment

 



Visit other IDG sites: