Yet another shuffle topic

[selyb]

After using Rockbox for 5 years +/-, I have evidently noticed every time that I would hear two tracks by the same artist separated by some other random track. On a side note, I have noticed this same thing in Foobar2000 :-p  Since I know a little about programming and I've written many scripts, I understand how when something doesn't 'seem' right, it really doesn't help on the programming side. I generated some numbers and all the evidence goes against how it feels when I'm listening.

Here's what I did: In rockbox, my shuffle option is always on. I added all tracks then saved the playlist. I then did reshuffle and saved to another playlist. Then I reshuffled again and saved to a third playlist. This playlist has 1734 tracks. In this collection, I have 345 uinque artists. 148 of these have only 1 track in the playlist. 130 have 2-7 tracks in the playlist.

Last night I wrote a short script which parses the playlist and checks for the same artist 1, 2, or 3 lines apart (1 meaning two consecutive entries, 3 meaning two lines separated by two other lines). In the shuffled playlist, The pattern matched 1 line apart 34 times, 2 lines apart 16 times, and 3 lines apart 17 times. The respective numbers for the reshuffled and rereshuffled playlists are 15, 5, 16, and 5, 17, 18. These numbers make more sense if you look at the linked logs. The zip file contains the three playlists (ANSI with DOS EOL), the output logs from the script, the vbscript I wrote to grab these stats and a stats.txt which shows how many tracks each artist has in this playlist.

The only problem I see with my tests is that nearly 1/3 of my playlist are one-track artists and my script checks for multiple tracks by the same artist. I may remove these entries and run these tests again in the near future.

Here is the zip file https://docs.google.com/file/d/0Byqa3Da0arZqb1J0b3NkS0FkZEU/edit?usp=sharing


Summary: When listening I often notice when I hear two tracks from the same artist close together, but after looking at the stats, it really doesn't seem to happen very often. In these 3 tests, of the 130 artists with 2-7 tracks, there were, on average, 5 instances where I had multiple tracks from the same artist within 3 playlist entries.

[gevaerts]

Looks interesting. I'd say this is all fairly natural, I've never heard anyone mentioning sudenly noticing that two consecutive tracks were *not* by the same artist, so that's obviously not very memorable :)

Now all we need is some proper statistical analysis that tells us how many of those events are actually expected...