[Appleman]

I have found iTunes / iPod while shuffling songs not that random
(Since this thread is moved, it seems one cannot post replies anymore, so that's why I started a new one).
I have made a Smart Playlist so you are sure you will hear all songs instead of same songs over and over.

Now how to post a pic here....

Anyway:

New Smart Playlist
Play count is 0
Play count is not 1
Live updating
(Reset Play Count for the first time)

Now you are sure all songs will be played before you hear a song a second time.

[yukon]

If it was entirely random, iTunes would be just as likely to keep playing the same song forever as play any other, we don't like that. iTunes uses a good algorithm to try to decide which song to play next. Unfortunatly it seems that there's only one algorithm, so often you see the same songs play after any given song, you can sometimes see that kind of pattern. I see it every once in a while, and can reproduce it.

Try using a playlist that selects songs at random, limit it by however many below your library total, and shuffle that. I found that by adding songs to the library or removing them, the algorithm would take that into account and change the list.

[Appleman]

Originally Posted by yukon

If it was entirely random, iTunes would be just as likely to keep playing the same song forever as play any other, we don't like that. iTunes uses a good algorithm to try to decide which song to play next. Unfortunatly it seems that there's only one algorithm, so often you see the same songs play after any given song, you can sometimes see that kind of pattern. I see it every once in a while, and can reproduce it.

Try using a playlist that selects songs at random, limit it by however many below your library total, and shuffle that. I found that by adding songs to the library or removing them, the algorithm would take that into account and change the list.

As I said, it does seem to have a certain patter, which I don't like.
While driviing for wo hours, 7 songs of the same CD passed by, while there is a choice of 7543 songs of 635 CD's. In two hours 7 of 7543 song from the same CD while the total should last 23.2 days!!
I understand random can mean theoretically it will simply play all songs in the order of the library, but I guess the algorithm will prevent that.
Does anyone know what kind of / which algorithm iTunes is using?

[Goldfinger]

Yep, I find this extremely frustrating. A bad algorithem sorta defies the point of having a +13000 song library when you always hear the same songs.

[Earth Mk. II]

I don't see why Apple would make a shuffle feature more complex than it has to be. I mean... it's random, and randomness does not imply a normal distribution of played songs.

As for the library repeating tracks, I made with a small (53 track) playlist and noticed something: The iTunes Library (and all playlists) shuffle without replacement. So long as iTunes is not interrupted after it begins playing, it will not repeat songs. Ever.

If, however, you cycle the shuffle option, or stop/quit iTunes, or switch to another playlist, the playlist or library will re-shuffle and it will forget about the last session.

now, it's possible that iTunes is giving the pseudorandom number generator a poor seed, and that would explain the reproducibility of the order in some cases. But I don't think the algorithm is biased in any way. Since computers are inherently deterministic and can never truly be random, this is likely.

If the play order seems to be non-random, I'd say that Apple is either using a poor pseudorandom number generator (or a poorly seeded pseudorandom number generator) or you're hallucinating and finding patterns that don't actually exist.

[Appleman]

Originally Posted by Earth Mk. II

or you're hallucinating ...

I doubt it

[Earth Mk. II]

Originally Posted by Appleman

I doubt it

Newsweek did an article on this a while ago too. Quoth the Apple engineers, "Random is random."

You may be perceiving a pattern, but there probably isn't one there at all.

[jamil5454]

iTunes could play the same song 3,000 times in a row, and that would still be considered random. Since random deals solely with chances, we'll never be able to prove something is totally random. We can only guess and assume.

[nonhuman]

Originally Posted by jamil5454

iTunes could play the same song 3,000 times in a row, and that would still be considered random. Since random deals solely with chances, we'll never be able to prove something is totally random. We can only guess and assume.

If you leave a playlist playing on shuffle for 100x the total length of the playlist, truely random play would give you approximately 100 plays per song with some, but not too much, variation.

If there was any significant variation it would be a pretty good indicator that the random number generator iTunes is using does a bad job at approximating true randomness.

[CaptainHaddock]

"So long as iTunes is not interrupted after it begins playing, it will not repeat songs. Ever."

I don't think my iTunes works that way. Sometimes I get the same song twice (or even three times) on party shuffle's list of upcoming songs. And I have about 3000 songs.

[Appleman]

Originally Posted by CaptainHaddock

I don't think my iTunes works that way. Sometimes I get the same song twice (or even three times) on party shuffle's list of upcoming songs.

That's where my Smart Playlist is coming in: although it does not prevent 10 songs from the same album being played, at least a song will never play twice.
Still the point is not so much that I have songs playing double, but after listening for a while, stop iTunes, and then start again, very often you hear the same songs again (specially in the beginning.
Probably after letting it go on for a couple of days the algorithm starts to understand I really like variation

[Gee4orce]

I think the issue here is that humans are very, very good at spotting patterns - we can't help it. I'm pretty sure if you did a mathematical analysis you'd find that iTunes is random, it's just that we are seeing patterns that are not there.

That said - I've once had iTune play exactly the same song twice in a row, out of a selection of over 5000.

[Appleman]

Originally Posted by Gee4orce

I think the issue here is that humans are very, very good at spotting patterns - we can't help it. I'm pretty sure if you did a mathematical analysis you'd find that iTunes is random, it's just that we are seeing patterns that are not there.

That said - I've once had iTune play exactly the same song twice in a row, out of a selection of over 5000.

If there would be a possibility to make a Smart Playlist that let iTunes play one song from a album and then does not touch that album anymore, like shuffling over albums, but only play one song per album, then more variety would be achieved. But shuffle over album seem to play the whole album before skipping to the next, so actually is doing what I don't want!

[CaptainHaddock]

That's where my Smart Playlist is coming in: although it does not prevent 10 songs from the same album being played, at least a song will never play twice.

I have a smart playlist that omits any songs I've heard in the last two weeks. The only weakness occurs when the same song makes it into my party shuffle list 2 or 3 times (before it gets played and removed from the smart playlist).

What we randomness aficionados really want is a pseud-random shuffle that (1) avoids playing songs from the same album or artist too close together, and (2) weights the randomness so that the least recently heard songs are more likely to be played soon.

[Millennium]

iTunes is random, in fact it's completely random (or as close as computers come, at any rate). That may be the problem, because sometimes people don't want complete randomness. You can take songs out of a playlist once they're played, for example, but then you're no longer completely random. You've introduced a pattern, however simple: all songs will be played before any are repeated.

[Earth Mk. II]

Originally Posted by CaptainHaddock

"So long as iTunes is not interrupted after it begins playing, it will not repeat songs. Ever."

I don't think my iTunes works that way. Sometimes I get the same song twice (or even three times) on party shuffle's list of upcoming songs. And I have about 3000 songs.

Party Shuffle is an entirely different beast. It's random with replacement, presumably so you don't run out of popular songs at a party. This is by design. You can also manually reorder Party Shuffle's playlist (you can't do that with the library or a normal playlist when the shuffle feature is active), and weight the song selection by a track's rating. Party Shuffle is not the shuffle function in question here.

The normal iTunes Library shuffles without replacement and is unweighted. It will eventually stop playing (or play through all songs then re-shuffle if the 'repeat' function is active).

Originally Posted by Millennium

You can take songs out of a playlist once they're played, for example, but then you're no longer completely random.

Sure it is! You've just reduced the set of possible events by one. That still allows for a random selection within the set of remaining songs.

It's true that non-played tracks will become more likely to be played next as each track is played (and subsequently removed from the set of available songs), but the selection procedure is still as random as a computer can get. Remember, the list of played songs is an ordered set. It's the order of the set that needs to be random, not the members.

[Millennium]

[QUOTE=Earth Mk. II]Party Shuffle is an entirely different beast. It's random with replacement, presumably so you don't run out of popular songs at a party. This is by design. You can also manually reorder Party Shuffle's playlist (you can't do that with the library or a normal playlist when the shuffle feature is active), and weight the song selection by a track's rating. Party Shuffle is not the shuffle function in question here.

The normal iTunes Library shuffles without replacement and is unweighted. It will eventually stop playing (or play through all songs then re-shuffle if the 'repeat' function is active).

Sure it is! You've just reduced the set of possible events by one. That still allows for a random selection within the set of remaining songs.

That doesn't matter; random playback functions over the set by definition. When you change the set in the middle of the function, you introduce a pattern.

It's true that non-played tracks will become more likely to be played next as each track is played (and subsequently removed from the set of available songs), but the selection procedure is still as random as a computer can get.

You just used the term 'more likely', which negates randomness. In a truly random setup, no song is more likely to play than any other, and this holds true at any point in time.

Remember, the list of played songs is an ordered set. It's the order of the set that needs to be random, not the members.

No, but the set you're picking from needs to stay fixed for as long as you're picking from it.

[Earth Mk. II]

Originally Posted by Millennium

That doesn't matter; random playback functions over the set by definition. When you change the set in the middle of the function, you introduce a pattern.

You just used the term 'more likely', which negates randomness. In a truly random setup, no song is more likely to play than any other, and this holds true at any point in time.

No, but the set you're picking from needs to stay fixed for as long as you're picking from it.

I will give you that the events are not independent. The selection of the next song is dependent upon the outcome of the previous selection, however, the selection of the next song is not caused by the outcome of the previous selection. There is no causal relationship between the two events of the current and next songs being chosen, as such, the events can still be random.

Simply saying that one event is dependent upon another event's outcome does not "negate randomness." The qualification that the sampling is being done without replacement does imply that the events are not independent, but that is all. Perhaps you could argue that sampling without replacement is "less random" than sampling with replacement, but that seems to be a semantical argument with little practical benefit, if any.

If I had 100 songs, each song has a probability of 1/100 to be chosen as the first song played. A song is selected, and removed from the set of available songs. Now, each remaining song has a probability of 1/99 to be chosen as the next song. In this way, the events are dependent on each other. However, this does not in any way determine which of the remaining songs will be played next. The selection is still random.

More succinctly: The shuffle feature reorders the played sequence of the playlist or Library into a random permutation of itself.