Yes, the iPod really is random
By Joshua Fruhlinger 
posted Feb 7th 2005 3:32AM
posted Feb 7th 2005 3:32AM
We've all heard it: "OMG! My iPod always plays Alicia Keys! It loves her as much as I do!" Apple's engineers swear that the shiny white music players actually do use pseudo-randomizing algorithms for shuffle play. The people at Newsweek were kind enough to verify this with Steve Jobs when they found that their iPod had an unhealthy obsession with Steely Dan (it happens). The conclusion? People love their iPods so much that they look for patterns in the pseudo-randomness. It's what we do, as humans.
When people start displaying anthropomorphic behaviours over their iPods it really is time to sell up and buy a desert island.
It's like the apes in 2001 who find the stick and are amazed at it's power... people are stupid.
this would be more amazing if the iPod actually played Steely Dan *even though* you own not one Steely Dan track... and never even put a Steely Dan track on it ... EVER. THAT would be freaky.
this is on par with people seeing faces in everything. or thinking that tarot cards or horoscopes or nostradamus quatrains in some way can accurately predict the future. people like random events and random formations to contain a deeper meaning with a deeper power behind them. they don't like the idea that in some cases... things really do just happen by pure chance and completely beyond ones control.
i presume the word 'random' is prefaced with 'pseudo-' because the algorythms by their inherent nature really aren't completely 100% random? in that... if they were you might *randomly* (though highly unlikely) hear the same song like 58 times in a row? and another song not once?
Hmmm, Steely Dan.
At least they're building some good taste into 'em.
Which song? Last Mall? Everything must go?
FM?
I actually thought the same thing for awhile, that my mini wasn't entirely random, so I did an easy check. Since iTunes shows how many times a song is played, I just plotted up a nice little bar chart and got more or less a bell curve. I was bored at work, what can I say?
I say bull....on my Ipod I have created playlist of maybe 100 songs each.
When I put the ipod in shuffle mode and play a playlist it will play song (a) then song (e) then song (t)
I then go and play a different playlist
plays song (a) then song (p) and then song (q)
come back and play the first playlist
it plays song (a) then song (z) then song (w)
it always plays that first bloody song of the playlist then the rest is random.
So, if your first song in a playlist is Steely Dan, you have to get through that one first and that is why you always hear it
this is on a Ipod with a 40 gig drive
and yes, I have a testing plan written out to follow and track it, so yes, I know what I am talking about.
From on-line dictionary...
Apophenia is the spontaneous perception of connections and meaningfulness of unrelated phenomena. The term was coined by K. Conrad in 1958
First of all, Stelly Dan is a dink. Second, Alicia Keys is a quadrapod. 3rd, there really isn't a definition for quadrapod - yet. I made it up 15 years ago. We all very well may be 'quadrapods'.
Pav said:
"I actually thought the same thing for awhile, that my mini wasn't entirely random, so I did an easy check. Since iTunes shows how many times a song is played, I just plotted up a nice little bar chart and got more or less a bell curve. I was bored at work, what can I say?"
Shouldn't you be getting more of a slightly wavy line? A bell curve shows normal distribution and/or the fact that the songs in the center of your bell curve are getting more play time than those sorrounding it(on the curve).
Actually, there does seem to be some flaws in randomization - as if they don't use a new seed each time, or that the next random is calculated somehow based on the files in the playlist, etc.
Very often I'll play the same album or playlist and get (nearly) the same order.
This is on a 3G iPod, and I haven't necessarily upgraded the firmware in awhile, so maybe it's since been fixed :)
I did a bar chart where I counted up the number of songs played once, twice, etc. In the end, I only had a handful of songs played once, one song played something like 12 times, but the majority were in the 6-8 range.
The perception that the iPod is not really random, is just that. In order to have randomness, some songs MUST be played more than others. If I'd done that chart and every song had played 6 or 7 times, that would have shown it's not random at all.
In terms of the ipod shuffle, randomness will be less random (if you know what i mean) than with a normal ipod due to the lack of an internal clock - which i presume the normal ipods use as a seed.
"First of all, Stelly Dan is a dink. Second, Alicia Keys is a quadrapod. 3rd, there really isn't a definition for quadrapod - yet. I made it up 15 years ago. We all very well may be 'quadrapods'."
I must be a unipod as I only have one ipod. Unless you are referring to some other "pod" that makes Alicia Keys a "quadrapod" :-)
Um Pav, you've got it wrong. If the Shuffle were truley random, than each song would have the same chance of getting played at any time. If you were to plot this over time, you would find an even distribution. Basically a flat line. If you are getting a bell shaped line (Which others are reporting), that means it is in fact not random.
Do a little experiment that is random, and chart it up and you will see a normal distribution (aka bell curve). I just did a very simple one to verify things for myself. Take 10 coins and number them. Pull one (without looking), mark down that you pulled it and return it to the rest of them. Keep doing the same thing and you will see a distribution that is not flat. Eventually, all of them should be chosen, but some will be chosen more often. Here are my results:
Coin # - # of pulls
1 2
2 5
3 3
4 0
5 3
6 2
7 1
8 2
9 4
10 5
Which leads to the following distribution:
0 +
1 +
2 +++
3 ++
4 +
5 ++
That is entierly random, or at least as random as you can make it, and you can already see that some are chosen more often than others and all have equal chance of being chosen. Increase the population to 1000 or more and pick a few thousand times and you will see a similar pattern emerge.
I don't know if anyone is still reading this but if you gave a million people a random playlist generator, it makes sense that there will be at least a few people that has lists genereated that don't appear random.
For what it's worth, I have a B.S. in computer science. My understanding is that computer algorithms are called pseudo-random because, although over a large number of trials, they approximate the same distribution as a truly random number generator would, they are completely repeatable, which a truly random generator would not be. I.e. given the same "seed" value, a pseudo-random-number generator will produce exactly the same sequence of "random" numbers each time.
Misconception #2: just because a pseudo-random-number generator produces a string of the same number doesn't mean it's not random. If you roll a perfect die 100 times, there is a small chance that it will come up six every time (1/6 ^ 100, if I remember my probability, but I got all C's and D's in math, so I could be wrong). People sometimes mistakenly believe that random sequences don't repeat. A random sequence is just as likely to repeat as it is to do anything else.
Misconception the third: A truly random sequence will produce a uniform distribution, which looks like a flat line. If you flip a coin or roll a die and one outcome comes up more than any other, it means that the coin or die is weighted and is not truly random. If you're tracking your iTunes play statistics, it might take a very large number of samples before the uniformity of the random distribution becomes apparent, depending on the nature of the pseudo-random-number generation algorithm.
What my distribution shows is that for the most part, the songs have been played 6-8 times, which is about as close to flat as you're going to get. Realize that I am plotting the number of songs chosen 1, 2, 3, 4, etc. times, not the number of times each song has been chosen. By that, for randomness, I should show that the majority of songs are played about the same number of times, in this case 7 times. But there are outliers that have played 10 times or 1 time. I believe we are talking about the same thing, just doing so differently. The way I look at it, if I pick any song in my playlist, it has probably been played 6-8 times. However, there is a small possibility that it could have been played once, or that it could have been played 10 times. Roll a die 60 times, and you will see that some numbers come up more often than others. Roll another 60 times, and chances are another number will come up more often. That is random. If each number came up exactly 10 times consistently, than something isn't right. A bell curve, the way I am using it, shows a more or less flat distribution for songs being chosen. Remember, I am not looking at individual songs, I am grouping songs based on the number of times they have been chosen and plotting that. I am quite certain of what I am saying since I have my statistics book open right next to me. I may suck at explaining it, but my reasoning is accurate.
Now, a way to prove that it is not, in fact, random would be if I removed the play count and over time exactly the same songs played the same number of times, but I suspect it would be different songs playing in that sweet spot of the curve.
But that's not the point. The randomization algorithm shouldn't be pseudorandom - it should actively seek out variety and do elimination play.
One potential outcome of random is real patterns and the iPod should take measures to avoid these particular outcomes because it's not what the listener wants, no matter what the statiticians say.
You know, I feel the name "iPod Shuffle" is deceptive if it is truly attempting to play the songs randomly. From my experience, in audio-playlist lingo, there is a difference between shuffle and random. Random will, theoretically, randomly choose any song on the list at any time, so you might easily get repeats. Shuffle, on the other hand, does a single-event random re-ordering of the list so each song plays once, but in a different order than before. For example, the difference between the "Party Shuffle" feature in iTunes and making a playlist and hitting the "Shuffle" button on it. The former will give you repeats. The latter won't until the playlist loops, or you re-shuffle. I prefer the latter, myself.
I've been a shuffle player since I got the first iPod. Last year I used my ADC account and wrote Apple saying "Wouldn't it be cool if there was a menu item to let you re-shuffle your play list, cause the darn thing won't re-shuffle your list until it hits the end or you load more music." The response I got back was immediate and obviously very practiced, and of course totally ignored my shuffle idea. Yep, mention their random number generator and they get all weird on you. Anyway, they added my requested feature (finally), didn't give me any credit for thinking it up (boo hoo, but I forgive them since they added it) and they tweaked the random number generator. Their number generator was always more of a brownian motion drunken walk than a true white noise. I think they did/do this to conserve power, if you don't make the ol' head in the hard disk move around so much you save juice. The best way I've found to shake things up is to put more music on the iPod.
well my ipod loves brazillian music. I have 2200 songs, of those maybe 7 or 10 CDs are Brazillian music, and well...I hear a Brazillian track about once every 5 or 7 tracks. You do the math.
Well, uhh, I only have about 180 songs on my iPod, around 30 are Metal (Hard Rock and all, which isn't my favorite genre) and I did notice that every 3 or 4 songs it played one of this 30, at the beginning it was ok, I thought that it was something temporary, I could get used to it. But after a month of constant lyrics about hell, the cruzades, lords, masters, and this repetitive, medieval, but yet catchy 15-minutes Dream Theater songs I created a Smart Playlist that excluded most of my metal songs. I feel much better now. At the gates and the walls of Monts?r. Blood on the stones of the citadel.
That's Iron Maiden, Monts?r.
Well, I don't know anything much about iPods, but my Dell DJ does some strange things with Shuffle on. I have 943 songs on it, which consists of all of my CDs plus some songs by Daughter Darling, Collide and Velvet Chain acquired from each band's website. An example of the oddities, obtained by clicking Next repeatedly while in Shuffle mode:
1. Aimee Mann - How Am I Different
2. Smashing Pumpkins - Rhinoceros
3. John Mayer - 3x5
4. Aimee Mann - Fall Of The World's Own Optimist
5. Emiliana Torrini - Wednesday's Child
6. Smashing Pumpkins - I Am One
7. John Mayer - No Such Thing
8. Filter - Hey Man Nice Shot
9. Emiliana Torrini - Fingertips
...and so on. Note that it seems to be repeating based on the album. For example, "Rhinoceros" and "I Am One" are from the same album by Smashing Pumpkins, and I have nearly every album by them. The same is true for each artist or band mentioned above. This is consistent across the entire player. It's most noticeable when I'm driving with it hooked up through the car stereo; I'm not looking at the display but rather I'm hearing the pattern.
The randomizing algorithms used in most audio-video devices are insultingly primitive and fool nobody. This is because the shuffle function has, until recently, been considered a trivial frill. Now that it is a major marketing feature, manufacturers ought to go back to their drawing boards and create a random function that is seeded by an event external to the device itself, e.g. the exact time it is turned on.
My shuffle is one day old, and I could swear it's repeating the same sequence.
Is it possible that it resets the seed for the random number generator every time it is switched off?
My iPod absolutely loves a song called Jack the Bear by Duke Ellington. Anytime you restart it or anytime you turn it on when it has been off for a long time, it starts with that song.