Bug 347058

Summary: Is the search in random on dynamic biassed towards similar tracks?
Product: [Applications] amarok Reporter: robert marshall <robert>
Component: Playlists/Dynamic PlaylistsAssignee: Amarok Bugs <amarok-bugs-null>
Status: RESOLVED NOT A BUG    
Severity: wishlist CC: ralf-engels, tuomas
Priority: NOR    
Version First Reported In: 2.8-git   
Target Milestone: 2.9   
Platform: Ubuntu   
OS: Linux   
Latest Commit: Version Fixed/Implemented In:
Sentry Crash Report:

Description robert marshall 2015-05-02 13:15:05 UTC 
I put something on the devel list months ago about this but didn't see any responses.

In dynamic playlist I have various 'and' conditions one of which has 
match all
   next track from album
   else: search for tracknumber:1
That seems to frequently give an album which is similar to the previous one (once there are no next tracks in the album) - maybe it's just a coincidence - but often I see a very long symphony immediately followed by another very long symphony - I would like the facility to say don't look for something similar.

For example, at the moment my  playlist has (provided by dynamic playlist)
a performance of the Schubert String Quintet (4 tracks)
another performance of the Schubert String Quintet (4 tracks)
an unrelated single track
another(!) performance of the Schubert String Quintet

These feel too closely related to be random  - the various other 'and' conditions don't limit the possible tracks from the collection that much. The 'and's limit to lastplay>7m and playcount:0 and that gives 3200 tracks, within those there are 16 tracks which are the String Quintets.

If there is the link to the prev album in the code somewhere, (I've not found it yet), I've like to be able to be able to break the connection in the dynamic playlist specification.

Reproducible: Always
Comment 1 Tuomas Nurmi 2024-07-12 23:04:48 UTC 
I've skimmed through the dynamic playlist code a bit, but haven't seen anything like that anywhere.

I guess this is at least partially explained psychological confirmation bias; additionally, this effect is probably intensified by the fact that long symphonies are, well, long, so they're being played for longer times even if they come up as often as shorter tracks. And, albums with longer tracks are likely to have less tracks on album, so a greater proportion of them match the tracknumber:1 too, I guess.