I have a collection of about 80k tracks. About 5% of these tracks actually have star ratings in Plex. Of those tracks that are rated, about 20% have ratings higher than 2 stars.
When I play a station in Plexamp (like “celebratory radio”), I notice that when a track that comes up is rated, it always has a rating of 3 stars or higher. That is, even though nearly 80% of the rated tracks are rated 2-stars, I never see 2-star rated tracks show up in stations that I play on plexamp.
Note that the behaviour for mixes seems to be different. I can get 2-star tracks in mixes.
This is not what I want. The star ratings meaning to me is below. (and of course this is individual. Everyone has a different scheme.) Basically I see tracks as “yuk”, “meh” and “banger”. I chose this scheme so I can have artist-scale, genre-scale and library-scale levels of “bangerness”.
1 star: Never want to hear this track again. (not many tracks)
2 stars: Happy for it to come up in a shuffle, but not a banger (about 80% of rated tracks)
3, 4, 5 stars: bangers, super bangers and mega bangers. I use these to build playlists. (The cream)
But it looks like Plexamp is seeing 2 stars as “I never want this to come up in radio but it can come up in mixes”.
I cannot find the exact impact of star ratings on radio documented, nor have I managed to find a way to configure how star ratings impact shuffles/radios. (I thought there was a way to turn it off, but can’t find it ATM.)
This kind of behaviour should be configurable, or at least documented so I can match my star ratings to the behaviour.
I just generated 15-20 play queues using mood radio. All but one of the queues had at least one 3+ star track and most had 2 or more. Not a single one had a 2-star track in it.
Out of 83.3k tracks, 3740 are rated 2 and about 20 are rated 1, and 1105 are rated 3,4 or 5.
That is indicative of a very strong bias agains 2-star tracks. (and about a 2-1 bias towards 3-star tracks over unrated tracks.)
This using the latest Android version of Plexamp and the latest-but 1 release of PMS.
Would it be possible to give some indication of the strength of the biases for: high-rated vs. unrated vs. low-rated in:
radio stations
mixes
smart shuffles
I have this queasy feeling that high-rated tracks are so strongly preferred over unrated tracks that once I have about 5-10% of my library rated such that it can appear in radio, that’s pretty-much all that I am going to hear on radio, negating radio as a tool for discovery.
So is it fair to say that 5/10 (i.e. 2.5 stars in the Plex interface?) gives a flat and unbiased preference for a track? And is that preference any different from that for an unrated track?
Is 5/10 in some way special? Looking at the one-star system, I think you mean below 5/10 and greater than or equal to 5/10?
In other words, on the scale of “Yuk”, “Meh”, “Yum” I seem to be seeing
1-4 == “Yuk”
unrated == “Meh”
5-10 == “Yum”
What is the actual strength of the bias? My experimentation suggests the bias is pretty strong - of the order of 5:1 relative preference between the levels?
Assume none of the tracks are “recently” played. (But what does “recent” even mean?)
