Fun with iTunes Shuffle Feature

i haven't tried iTunes shuffle much but I've noticed over the years that computer programs often have a hard time w/ true randomness.

 

iTunes shuffle has always been a little silly...I maintained that it had a brain for many versions, but now it has a little slider so you can set how random it really is...I was right all along.

 

Or, are they just really good at it? I'm no statistics expert, but I do know that humans are really bad at "randomly" picking things from a list, even though we think we would be good at it. Perhaps it's not perfect, as Mr. Durden says over half of his songs haven't played yet...but maybe it's better than we give it credit for.

What does everyone else think?

 

It uses a random number generator with a seed variable. That seed variable is usually the time, but the shuffle doesn't have a clock so it has to use something else. I bet that slider is just a number to be input in the random number generator.

BEN

 

Ummm...he was taking about the iTunes Shuffle feature, not the iPod Shuffle.

 

Not true with a lot of things in Windows (bugs, viruses, spyware, etc.)

 

Yeah, i've noticed that itunes will only shuffle to songs I have listened to... or really hate. There are certain artists i will NEVER, EVER hear, and others that I'll hit within 3 clicks... and this is in a 10,000 song, 1,000 artist library.

 

If you play songs from your playlist on random, I believe it chooses your 5 star songs, and songs with the highest play count more often than the songs that have never been played all the way through.

 

i think the shuffle function will function correctly, if one uses the ratings correctly.

 

I hate how iTunes only randomizes songs the instant you click on the "shuffle" button... From then on, iTunes has a set order in which it will play the songs... So it doesn't use a time seed variable or whatever... You have to re-shuffle songs... Gah!

 

Randomness

There's some confusion here. Apple is probably using a 'pseudo-random' number generator- which although deterministic- it is practically indistinguishable from a real random sequence (such as tossing coins).

It is natural, but quite mistaken to think that with a library of 1000 songs and performing say 5000 shuffles that you should get fairly even sampling of each song. However, there's an inherent noise with random numbers (goes as 1/sqrt(N)) whereby you don't get a smooth even distribution unless you run for many many steps, maybe 100,000. For only a few thousand shuffles, random numbers will almost certainly hit several songs multiple times, whereas other songs will not be hit at all. The fact that you observe 'spotty' selection of tunes is not inconsistent with the fact that a random number generator is being used- in fact- it is exactly what it should be like!

This is the basis of 'Monte-Carlo' sampling methods used by physicists and other scientists. Wikipedia will tell you all.

 

The randomness really breaks down pulling from a fixed sized list.

I have a list of songs that meet certain criteria. I haven't listened to them in 2 weeks, they are rated 4 or better. Limited to 200 songs by least played. When I hit play higher rated songs more often it plopped the same song down 10 times in a row.