What's Wrong with the Shuffle Algorithm?

If it's using a random number generator then those sort of things happen.

It's just like lotteries - some numbers come up all the time, other far less frequently.

AC

Hi,

Not sure if this will help, but it seemed to help in my case. I was noticing a bunch of repeats and/or playlists "sticking" to a genre or such. What I did was to turn off the shuffle button on the playlist(s) in iTunes, then re-syncing the ATV. It seems to go a lot deeper into playlists now when on shuffle from the ATV. Worth a try?

Nah, CDs killed off hi-fi...

Ac

Great article mind you, thanks for the link.

AC

This is an old problem that goes back to the original iPod. Such algorithms aren't actually true random number generators, they use the device's clock to pick numbers, thus are limited by the clock cycle.

It also appears to me that the song selection is weighted by how many times a particular song has played, which means that the more often a song is played, the more likely it is to be selected again. It essentially creates its own favorites list.

The shuffle algorithm is definitely in need of revision.

wd5
Maybe this is a solution for you:
I made a smart playlist, including all songs, which looks at the playcount. A song is only included in the playlist when its playcount is 0. If you play this playlist on the Apple TV in shuffle mode each song will only be played once.

Regards,
Marc

Message was edited by: Marc Hofstede

Just a bit of useless info,

The only true rnadom event is the time between radioactive decay.

But iTunes / iPod doesn't help.
If you shuffle albums, a random track is selected and the rest of its album is played through to the end. So your randomness tends to favour the final tracks on an album.
If you shuffle tracks, you should get fairly good randomness. Not sure what it uses to determine the next track number, but you would need billions of tracks to throw out even the simplest of seeded pseudo-random number generators, ie. using the current track number to generate the next.

Are all the tracks selected, ie. have the little tick next to the track name in iTunes ?

Check out how the iPOD killed high fidelity at http://www.rollingstone.com/news/story/17777619/thedeath_of_highfidelity.

You're confused between two completely different processes, both of which are called compression. What the article is about is dynamic range compression: reducing the difference between the softer and louder parts of the music, in this case specifically by raising the loudness of the softer parts of the music to make the whole thing sound louder.

MP3 and AAC compression, as used in iTunes, doesn't do that at all: what it does is to reduce the data size of the music by both using standard lossless data compression methods and by taking advantage of psycho-acoustics, in that some sounds will mask (make inaudible) others if both are present at the same time. How noticeable MP3 compression is depends on the nature of the music being compressed and how severely it is compressed.

With regard to how the music sounds, MP3 and AAC compression aren't supposed to make the music sound different (any difference is considered a flaw in the process). Dynamic Range compression, on the other hand, IS supposed to make the music sound different: that's the whole point of doing it.

By introducing MP3 compression into the discussion, the article only confused matters. Dynamic range compression is aimed mainly at radio airplay: in general, the louder a song, the more it tends to capture listener's attention. This causes radio stations to compress everything they air in order to make their station louder and more attention-grabbing, and it causes producers to compress their music in order to make it louder and more attention-grabbing. These loudness wars began long before the iPod appeared on the scene and aren't being done because of it. Further, ripping all your music losslessly does nothing to fix it. Lossless compression may ensure that you don't make the music sound worse by introducing MP3 compression artifacts, but it isn't going to have any effect on problems caused by dynamic range compression in the original recording.