- Introduction
- Say You, Say What?
Say You, Say What?
Using Excel, we collated the results and counted the number of times that each song and artist appeared in the playlists iTunes had generated.
All things being equal (and random), one would expect that a field of 1300 song slots would provide enough opportunities for each of the 56 artists to be equally represented. However, this was not the case. In fact, even though each specifically chosen had five songs in the library, there was a large discrepancy between the most popular and least popular artists.
Lionel Richie (Universal) proved to be iTunes' most popular artist, appearing 59 times all told, for an average of 1.475 times per possible playlist (or TPP, an objective measure reflecting the fact that iTunes-purchased songs were available to iTunes during creation of 40 playlists while CD-ripped songs could only have appeared on 20 playlists).
Red Hot Chili Peppers and The Veronicas tied for second place (55 times / 1.375 TPP) with Keane and Robbie Williams (53/1.325 TPP), Eskimo Joe (52/1.3 TPP), Good Charlotte (51/1.275 TPP) and Grinspoon (50/1.25) all appearing at least 50 times in our playlists. Looking further down the list, however, a curious trend appeared. Artists whose five songs were bought from iTunes were consistently more likely to appear on the random playlists than those whose songs were ripped from CD. Each of the top 15 artists, by number of songs played and songs per playlist, was bought from iTunes; those ripped from CD were far less likely to rate.
Songs by Def Leppard, the most frequently-played artist from CD, were chosen 24 times in 20 possible playlists for a TPP of 1.2 -- but the rate of selection for other CD artists quickly dropped off: Bon Jovi (21/2 TPP), Creed (20/1 TPP), and Gloria Estefan (18/0.9 TPP) were all slightly ahead of Andrew Lloyd Webber, the Bee Gees, Dido, Erasure, Jackson Browne, Maroon 5 and Mötley Crüe, all of which were played 17 times for a TPP of 0.85.
The least frequently played artists were all those whose songs were taken from CD, with the bottom of the rung inhabited by Kate Bush (12/0.6 TPP), Anderson Bruford Wakeman Howe (11/0.55 TPP), and Christina Aguilera and Oasis (10/0.5 TPP).
When the artists with just one song were factored in, things got even more interesting. Smack That -- a current hit by Akon and Eminem -- was played 17 times, which was the mean for the artists with five songs in the iTunes Library. Its TPP was 0.425, a frequency that translates to 2.125 if treated as though there were five copies of the song (we'll call that ratio TPP5). In other words, we were 1.44 times as likely to hear Smack That than any song by Lionel Richie, even though Lionel Richie had five songs in the iTunes Library.
Also played proportionally more frequently than five-song artists were Gwen Stefani's The Sweet Escape and Hinder's Lips Of An Angel (14/1.75 TPP5), Justin Timberlake's What Goes Around/Comes Around (13/1.625 TPP5), and Daughtry's It's Not Over (12/1.5 TPP5). Beyoncé's Irreplaceable, Fall Out Boy's This Ain't A Scene, It's An Arms Race, Gym Class Heroes' Cupid's Chokehold, Nelly Furtado's Say It Right (11/1.375 TPP5) all had as many playlist appearances as songs from the Red Hot Chili Peppers and The Veronicas -- even though the latter artists should presumably have been five times more common.
Like this article? Click below to send it to your mobile for free!
general84
08/03/2007 09:23 PM
There are no calculations to show whether these findings were statistically significant over the dataset, there is merely conjecture while showing raw results. http://en.wikipedia.org/wiki/Statistical_significance
Report offensive content
D_Money_16
09/03/2007 11:42 AM
They should analyze it using the Chai squared analysis to determine if it is truly outside statistical variation or just chance.
Report offensive content
morbo
09/03/2007 11:55 AM
From just glancing at this, it looks consistent with most types of LCGs. In fact, these are pretty close, in terms of frequency, for all categories. A 3% difference is negligible. Random does not mean a bell curve, sorry C-NET. This is amateur statistics at best.
Report offensive content
iopred
09/03/2007 12:09 PM
This is rather ridiculous, what do record companies have to gain from you listening to songs you own more than other songs you own?
Report offensive content
dchris
09/03/2007 12:16 PM
Like others have said, this study really isn't properly done at all. Your sample sizes are too small, you didn't replicate your experiment, and you falsely draw conclusions like, "Random is relative," -- of course this is true if you've only done the experiment once! Flipping a coin 10 times, getting 8 tails, and then announcing that the coin is unfair is really bad statistical analysis. Also, you spelled "revelations" incorrectly in the sentence "...but the relevations in our study certainly lend weight..."
Report offensive content
dogarms
09/03/2007 12:17 PM
I just wanted to chip in that there is actually a slider in iTunes (under 'playback') to control the likelihood of hearing the same artist or album in a row. So if you like, you can vary this percieved 'non-random' randomness. But really, random is random. Humans look for pattens in random data that aren't there all the time.
Report offensive content
random()
09/03/2007 12:19 PM
Apple actually changed iTunes' randomizer because people were complaining about the same kind of things. There are several factors to take into account. First, most people don't have homogenous libraries. For instance, I do have a lot of songs from "Hawksley Workman," so yeah, I hear him more than most of the other artists in my library. Then, a random algorithm is not perfect, and cannot be "perfect." And then, we often "notice" the song we know or like best and are then lead to believe that some songs are played more often than the others. iTunes now lets you specify the amount of "randomness" you want. For instance, my iTunes is setup so that there's a higher chance to have two songs from the same artist in a row. That's what Apple introduced for people who thought iTunes wasn't random at all. If you want, you can ask iTunes to never play the same artist twice in a row. Last but not least, the other comments could not be truer. This study is utterly useless and shows a total lack of understanding about statistics and randomizers.
Report offensive content
tor
09/03/2007 12:43 PM
No Digg! Use some stats first
Report offensive content
spidermedic
09/03/2007 12:43 PM
To really test is you should have had copies of the songs in both formats. Then you could have done the same tests using the same songs, just in the opposite format (ripped vs iTunes store) If the iTunes store copies were still played more, you'd be onto something.
Report offensive content
jmc
09/03/2007 01:12 PM
Shame on you, CNET. Learn some statistics and refrain from doing sophomoric analyses.
Report offensive content
Gelf
09/03/2007 01:22 PM
Humans find patterns in randomness. It would be more surprising if you could find no pattern at all. You have already invented a conspiracy surrounding the music's publishing label, completely oblivious to the fact that there is no evidence iTunes even knows a track's label. You might as well look for a pattern of artist hair colors. You'd find one, by the way, using your test protocol. One good way to check yourself would be to run the test again now that you have committed yourself, in print, to certain alleged preferences in iTunes. Stop looking for just any pattern you may happen to stumble upon and try to reproduce those specific ones you've already claimed. You will likely surprise yourself. You will see other "patterns," but not the precise behavior you think you have already identified.
Report offensive content
glowingsoal
09/03/2007 01:45 PM
A lot of people have negatively commented on the accuracy of these findings, however the results are extremely accurate in the case of my itunes library. I have questioned the randomness of shuffle for awhile now and this totally supports my assumptions. I use shuffle often and have seen that Radiohead (EMI) and Christopher O'Riley (itunes purchased) appears at least once out of every 10 songs, while certain local bands are never played.
Report offensive content
rmk
09/03/2007 01:58 PM
Random is a hard concept to understand. Flipping a coin is a random event, but it's possible to flip heads a million times in a row. Ever play a board game, but just can't roll a 2 to get to the end? Never rolling a 2 doesn't mean the roll wasn't random. Pattern's seemingly occur in random events all the time - just look at the stars. How many shapes can you pick out? Big bang says star locations are in random locations, but we see patterns in them anyways.
Report offensive content
kevinmichaelkeating
09/03/2007 02:03 PM
@glowingsoal: When shuffling, iTunes simply re-organizes the entire playlist being shuffled, which means that EVERY song gets played once, until ALL the songs have been played, and then they are rearranged again. That's not true random, sure, but it does keep songs from playing twice. And ultimately, the solution to artists annoyingly playing multiple times at the expense of others is simple...delete them or remove them from your playlists.
Report offensive content
Jackattack
09/03/2007 02:41 PM
It sounded like those individual songs you loaded into iTunes came last. It would also make sense for apple to randomize your music weighted more heavily with the songs you most recently put in your library since it's probably what you want to listen too. Oh well, just my two cents.
Report offensive content
Davidlow
09/03/2007 02:45 PM
Serious suggestion for C-net: Get a statistician to perform an analysis. Really, that would be great. I've read other analyses of the iTunes randomness feature before, and just like this one they always reflect the same common misconceptions about statistical randomness. To finally put this matter to bed, someone should do it right.
Report offensive content
btp
09/03/2007 03:11 PM
even if this is true (it may or may not be), there are other ways to randomize your playlist. have it play through the songs in alphabetical order by song title. while it isnt exactly random, it is random enough so you dont get the same artist all the time (unless all their songs start with 1 letter).
Report offensive content
pianoroy
09/03/2007 04:59 PM
Please do a real statistical analysis. I don't see any standard deviation, mean, or distribution function...
Report offensive content
Mike
09/03/2007 05:08 PM
In the context of computing systems, "random" means "not predictable". It does not mean "well distributed".
Report offensive content
Dr Hans Zarkov
09/03/2007 05:46 PM
>Spidermedic and gelf are thinking the right way about this. These are some compelling initial results, but you need an independent replication. >Statistical analysis is (of course) important, but it's only going to tell you what the probability is that these results occurred by *random chance*; it's not going to help prove or disprove your theory. The current theory is that tracks by some publishers are played more than tracks from other publishers, right? But an equally plausible hypothesis from these data is that some numbers are 'sticky' - the random number generator likes them more than others, for some reason, and picks tracks tagged with those numbers. >How could we possibly differentiate between these two theories? I think that if you wiped your drive and re-downloaded the tracks in a *different* order and found a similar bias then that would be compelling evidence that something about the tracks, rather than the order in which they were loaded into iTunes, is responsible. >Then you're going to have to convert everything to MP3, wipe the drive again and put all the tracks back on. If it *loses* the bias now - if everything is basically random -then I think you've got a great story on your hands!
Report offensive content
Darabbit
09/03/2007 06:12 PM
Perhaps a nuclear powered Ipod could use radioactive decay to truly randomize the playlist. Another handy test here would be to create some MP3s of your own (a set of exact duplicates, perhaps 3 minutes of noise, except for 'artist' 'title' etc) and a set of varied tracks. This would be good control data.
Report offensive content
anonomos
09/03/2007 06:21 PM
If you wanted to make any arguments about how accurate the random number generator is why did you not use actual statistical equations, instead of making up your own terminology and make inferences with the standards that have been in place for a long time. Note: the math for doing it properly is easy with sample size of 30 of 25 songs or 30 of 40 songs.
Report offensive content
pintong
09/03/2007 07:02 PM
You ended with the question "Have you noticed any patterns in iTunes' randomiser?" Yes, actually. It seems that sometimes iTunes will fall into a loop when on shuffle. I first noticed this when doing a "post lyrics from the next 15 songs on shuffle" game. I noticed that as I approached song 15, the song order repeated. I verified this by continuing to skip forward until it looped again. This has happened to me a couple more times since. My best guess is it uses something in the song's meta to generate the "random" value for the next track. It might also combine it with info from the last track played, thus creating a loop when conditions are correct. This occurred under iTunes 5 and 6 (I believe). It has yet to happen under iTunes 7.
Report offensive content
bob
09/03/2007 08:44 PM
Seriously, guys, we want a better statistical analysis. On the other hand, what do I care? I'm about to switch away from iTunes because its menus are too damn slow.
Report offensive content
tonyskilling.blogspot.com
09/03/2007 08:46 PM
Still trying to find the "Page 2" link :|
Report offensive content
stephaneL
09/03/2007 10:10 PM
nex step: disassembling the random code?
Report offensive content
Em
09/03/2007 11:14 PM
I recommend a reading of the randomization chapter in Numerical Recipes in C. If you read that, you'll see an excellent explanation of the fact that computer-based randomizers (which are just algorithms underneath) can couple in a decidedly non-random way to the application. This will result in some decidedly non-random behaving applications if the random number generator hasn't been chosen properly. If such is the case here, it might be fairly easy to think up some deep dark reason explaining the non random behavior.
Report offensive content
drgnmstrnik
10/03/2007 12:31 AM
It seems pretty clear from your results that the ONLY thing this indicates is the randomness is based more on ARTISTS rather than songs. Additionally I have to go and agree with the first comment that the deviations you have found are in no way statistically significant, so by all means, yes they are random.
Report offensive content
dbo
10/03/2007 04:38 AM
Why would the record companies pay to manipulate your playlist? They've already got your money! They don't care how many times you listen to something, just that you fork over the cash! So no profit motive for the record companies. Why would apple do it? I'm not sure but I bet it has something to do with the fact that steve jobs is compensating for something...
Report offensive content
dbo
10/03/2007 04:43 AM
Why would the record companies pay to manipulate your playlist? They've already got your money! They don't care how many times you listen to something, just that you fork over the cash! So no profit motive for the record companies. Why would apple do it? I'm not sure but I bet it has something to do with the fact that steve jobs is compensating for something...
Report offensive content
john i
10/03/2007 06:02 AM
Ditto on this being based on an unscientific understanding of stats. That said, I use smart playlists to disallow any song that's been recently played and weighting songs with low or no play count higher. This forces iTunes to play songs it hasn't.
Report offensive content
karinova
10/03/2007 06:49 AM
Interesting. But ultimately meaningless, because the only way to assess whether the randomizer algorithm is "unrandom" is to examine its parameters-- not its output.
Report offensive content
Garth
10/03/2007 08:56 AM
This is an ignorant and useless article. The author appears to understand neither commonly accepted statistical methods for evaluating randomness nor the general nature of random data itself. Highly misleading.
Report offensive content
paul
10/03/2007 12:27 PM
has no-one seen "rosenscantz and guildenstern are dead"? randomness constantly throws up unlikely patterns and surprising repetitions. That's the nature of randomness. That said my itunes library does surprise me. two mornings leaving the house for work wearing my ipod shuffle i heard the same playlist - as well as i remember pretty much the same songs. now, there has to be a procedural explanation for that! Then again, i wouldn't bet against it.
Report offensive content
guest
10/03/2007 04:00 PM
the red hot chili peppers get beaten to death on my ipod too. I smell retaliation.
Report offensive content
walto
10/03/2007 10:50 PM
I've actually noticed the same sequence of songs being played if I quit and start up iTunes again and hit play without changing any of the settings. I've also noticed that from time to time, for whatever reason, iTunes will seemingly "feature" an artist and keep playing their songs - which randomly should rarely if ever happen (even though it's possible) because I don't listen to playlists, I listen to my entire library set to random and I have over 3500 songs. As a user's tip, press the "random" button while holding down the option key to rejigger the order of play. I do it all the time and it definitely helps!
Report offensive content
Gavin
11/03/2007 05:47 PM
"Four songs -- Christina Aguilera's At Last, Creed's What's This Life For, Crowded House's World Where You Live and Led Zeppelin's Nobody's Fault But Mine -- were in the iTunes Library but were not chosen for any of the 40 playlists generated during this exercise." This sounds like a "value added" service to stop you listening to bad music
Report offensive content
P-Dub
12/03/2007 11:01 AM
This is so completely bogus... starting at the title. The author of this "research" has no clue of what "random" even means. The second item in the "summary".... "On average, one would expect each song to appear on 6.5 playlists." is the key to while this whole idea is meaningless. RANDOM is not AVERAGE. Random means that the odds of a coin flip resulting in tails is the SAME no matter how many times you flip it... and 20 tails in a row is just as "random" as any other result.
Report offensive content
Kris
12/03/2007 04:55 PM
As many people already pointed out, this story lacks any credibility. The author should've picked up any book about statistics first and read about "statistical hypothesis testing". It is disappointing to see that someone tries to do something he has absolutely no clue about and later on this story gets published.
Report offensive content
the dude
13/03/2007 10:31 AM
you're all missing the point -- that random playback is a very non-musical and rather droll way to listen to music in the first place. who cares if it's done truly randomly or not - the art of a good play sequence, album or collection of music to listen to is something that apple is no good at, but then they don't claim to be either. they sell hardware for a living.
Report offensive content
Paul K
14/03/2007 06:23 PM
Staggeringly bad math. Making playlists of the iTunes Music store only songs and then purchases + mp3s on the other playlists confuses the analysis and makes the "TPP" metric uterly meaningless. In fact, purchased songs should appear in an average of 9.75 playlists and mp3s in an 3.25 playlists each. The expected mean number of appearances of 5-song purchased artists is 48.75 which is very close to the median of those 16 artists -- which implies little to no skew. The expected number of purchased songs appearing in no playlists is 0 while 5.6 mp3s are expected to never appear in a playlist -- close to 4. With mp3 only 5-song artists expected to appear an average of 16.25 times and individual purchased songs an average of 9.75 times, then random variation means that it is reasonable that some songs will appear more often then some artist sets. In summmary: the data presented (poorly) in this article indicates that the most reasonable interpretation is that iTunes shuffling behaves randomly and suspicions otherwise have no credence.
Report offensive content
PCPete
15/03/2007 12:42 PM
If you repeat the test another 100 or so times, taking into account others' suggestions re copy order, format, and so on... then you might get a truly randomly representative result. (I hate maths). There's an elephant in the room here : how are the tracks being selected and weighted, if there is such a process going on? If the metadata is being used, it should be fairly simple to swap metadata between tracks and see if the reported patterns follow the track or the metadata. There are also more subtle ways of encoding such information in great big binary files, like steganographic manipulation, which would be next to impossible to determine without pulling apart the code, as stephanet suggested. Next article: How the pseudorandom iPod shuffle feature is abused by the Recording Companies.
Report offensive content
jez
17/03/2007 09:35 AM
One assumes in all this that the Smart Shuffle under playback preferences was precisely centred - the slightest movement either way would skew all results, since it's designed to do precisely what you're investigating - play clusters of artist/album tracks...
Report offensive content
JMJ
30/03/2007 06:12 PM
The reason a label might want to get their songs played more is so that the listener decides that they like whatever particular style of music or artists that the label is associated with because he's hearing it so much. Then he'll be more likely to make another purchase from that label's repertoire.
Report offensive content
tk
15/05/2007 08:34 PM
This might be irrelevant, but somehow when I have a set playlist and I start with one song, the order of the following songs are always the same. For example, I start with Breathing - LifeHouse, it always goes to Nothing In my Way by Keane... Is there anyway I can prevent this from happening, cos it's bugging me.
Report offensive content
NB
17/05/2007 05:00 AM
I've noticed that when I set my Nano to shuffle, the beat/pace/type of songs that follow are set by the first song that plays. i.e. if James Taylor is first, then I may get a country song, followed by a folk song, etc. Or if I start with a hi-hop song, it will follow with an alternative that's got a heavy beat, and so forth. I wondered if there was an algorithm that determined the "random" order. Thoughts??
Report offensive content
Plugh
28/06/2007 06:01 PM
I can't speak much to the statistical validity of my observation, but in the 7 months I've been using a smart playlist that draws a random 16 hours from my music library of 3400 songs, there are 400 songs that have never been played, and I listen to itunes 8-12 hours a day. A truly random generation of the playlist would theoretically have played all 10 days of my music by now, but this is obviously not the case.
Report offensive content
samesong
29/07/2007 04:55 PM
not sure why labels would care about how often you actually listen to a song after you've already paid for it... I've always felt like itunes on random only goes through about 20% of my library, mostly the same songs.
Report offensive content
The_Cynic
21/08/2007 05:59 PM
Statistics. For all the people that whinged about the validity of the stats or the method by which the info was gathered need to realise that it's just some numbers that don't mean much and comments like "amateur statistics at best" are really clever mate. If this is so poorly done then do your own study. And ask yourself, why would I get worked up over a fun little analysis? Calm down and get a real job. In terms of the randomness of the playlists, I have some songs that repeat every 100 or so and some that I don't hear for weeks at a time... so I certainly think that there is cause for study to be done into this.
Report offensive content
mdeck
03/09/2007 06:37 AM
I have no music from iTunes (I prefer having a CD to go back to in the [inevtiable] event that I lose my hard drive). I only play by albums -- if I don't like a particular song, I remove it from rotation. I shuffle on albums, not by songs. The way my iPods play the music now is just plain not at all random. It seems to have two points in the alphabet and it'll sort of randomize around those points -- say a Beatles album, then Stone Temple Pilots; then it'll go back to the B's, maybe a different Beatles album or Beck (as far as I've observed, it's almost always later in the alphabet by artist's name), then back to the S's. Eventually, it'll get to the C's and the T's, but it takes a while, and rarely plays any artist outside of the two pointers. And if I resync, it starts all over -- I get repeated albums frequently. Sure, the songs are somewhat random, but the albums are not, and the artists are rather predictable. I just want it to pay attention to the other albums in my collection. Smart playlist using "least recently played" doesn't help for two reasons: (1) I have multiple iPODs (sync'd only when I get new music), and (2) it doesn't allow me to select by album. I've had the smart shuffle slider all the way toward "less likely (to play songs from the same artist/album)" since I got the iPod (although I didn't notice the "songs" vs. "albums" radio button -- I just switched that to "albums," we'll see if gets any closer to random on albums). I want an option to play the music on my iPod(s) the same way I played it when I had only (gasp!) actual hardcopy CDs: Play every album once, in pretty much random order (chosen from the "unplayed" side of my CD rack; never two albums from the same artist within 10 albums), until every album in my whole collection has been played once (moving it to the "played" side of the rack afterward); then start over with the whole bunch (when "unplayed" album count reaches zero, "played" list becomes "unplayed"). [People (especially my siblings, and I'm sure readers here) think I'm crazy/anal retentive, but there is not an album in my collection that I don't know and love.] And I know my request isn't for true randomness, but since moving to iPod, I have no way to make sure I listen to everything once through before starting over. I miss music that I haven't heard in the three years since moving to iPod.
Report offensive content
2fs
20/10/2007 01:53 PM
Currently, I have a playlist with about 125 songs. Almost all of the songs are released by indie labels. There are only a handful of repeated artists. The purpose of the playlist is that I'm winnowing it down to make a year-end mix. I'm playing it in shuffle and, if I don't care as much for a song, I'll delete it from the playlist; otherwise I'll keep it. If I've already heard the song (obviously, this refers to songs I've decided to keep), I'll skip it and go to the next song the shuffle function selects. From this 120-song playlist, there are five to seven songs that have come up multiple times. A couple have come up three, even four times. After observing this, I made a query amongst some friends about this seemingly non-random behavior; immediately after posting that e-mail, another one of those 5-7 songs came up, again. I again skipped it; the next track was another of those 5-7 songs. After another few songs, yet another of those 5-7 songs came up, again. This took place over a period of about half an hour. That is, over half an hour, three tracks came up twice, all three of which had previously come up multiple times such that I noticed their recurrence. This study doesn't address the issue of how iTunes behaves if you skip a track - but the numbers I describe above suggest that once its algorithms select a track, if that track isn't played to completion, it's likelier to come up again. Or maybe my iTunes just has a jones for David Kilgour's "Sun of God" (one of those 5-7 songs: it came up three times during that half-hour period).
Report offensive content
Blah
06/02/2008 10:00 PM
Seriously... Some people have WAY to much time on their hands!!!
Report offensive content
irish`1
14/04/2008 01:56 AM
out of my 3000 song playlist eminem plays 1 out of 5 times. There are around 50 songs but that is still way too much. I had never even listened to the cd before I made the playlist. I can't figure out why it keeps playing it.I have not rated it listened to it or bought from itunes!
Report offensive content
miketheTexan
07/05/2008 10:23 AM
My iPod has about 5500 songs on it. Kryptonit by 3 Doors down has played 36 times since Sep 06, when I got the iPod. I've never once directly played this song, it always comes up on random. In fact, their songs occupy 9 of the top 10 slots, and those all come from random plays. What's up with that?
Report offensive content
jakehotep
15/06/2008 10:43 PM
The real point here is that NO ONE seems to want a randomizer, just a program that reads how many times each song is played, then distributes them more equally. Listening, Apple?
Report offensive content
woodseatsdj
21/06/2008 09:02 AM
If true randomness happens, it is quite plausable that the same song could be heard 6 times in a row - or never. A human might not think that random but the maths works. Try putting last weeks winning lottery numbers on and expect them to come up - it's just as likely - get yer ed 'round it. By the way, I heard that Psophometeric was spelled thus.
Report offensive content
swedeo
02/08/2008 10:45 AM
Heck Yes I think it should be studied more.. I have like 29000 songs on my 60 gig Ipod and yet I hear the same songs all the time.. and very rarely do I hear a change. I always have it on shuffle so I get the most out of the music I have. Apparently not so much
Report offensive content