How Do Star Ratings Work Across Multiple Song Of The Same Artist?

General Discussions Plex Features
1

Good morning,

I came across something not previously noticed when using the new (awesome) Plex Amp 3.0 …

I heavily rate songs in my library, and make smart playlists based on those ratings (songs with 5 stars, not played within the last 30 days … as an example).

I noticed when looking at a playlist contents lately that if I have rated lets say, Whole Lotta Love with 5 stars off of Led Zeppelin II, then Whole Lotta Love from lets say Mothership (A greatest Hits collection by Led Zeppelin) also has 5 stars, and BOTH appear in the smart playlist. The Mothership version has zero plays, so I could not have had the opportunity to rate that particular track off the Mothership collection.

Here is the question … When you rate a song by an artist. Does it apply that rating to all versions of that song from different albums the song appears on by the same artist? The ratings criteria is by song, rather than by specific track?

Thanks for reading

Stay home and stay safe / have a great day!

EB

1 Like
2

I was just about to write a post with the same observation. It does do what you describe. If you have two versions of the same track, by the same artist, on different albums, and you change the rating on one, the other is changed as well.

Whether this is by design or a bug, I can’t say. It seems counter-intuitive to me, and counter-productive. As you describe, if you have multiple versions of the same track (one is re-mastered, or an extended cut, for example), I would think you would only want the version you rated to appear in a smart playlist.

I tested something out. If you have tracks with identical titles, but by different artists, this does not occur (I have “All or Nothing at All,” one version by Diana Krall and one by Al Jarreau). Rating one track does not affect the identically-named track by a different artist.

3

Yes, I agree with everything you said and have observed the same. If this is by design, perhaps the devs can review it. I understand not all people have a large library of music and rating a song would be sufficient, BUT, I happen to have at least 9 versions of Whole Lotta Love by Led Zeppelin.

The way I have a particular smart playlist set up is that 5 stars and not played in the last 60 days. But what winds up happening is that while listening to this smart playlist I hear Whole Lotta Love 3 or 4 times, which to me defeats the playlist purpose and the criteria. Renders it almost useless. Besides, the live versions is clearly the superior version of the song, and the studio version has a certain classic appeal to it, but both versions I consider 5 stars worthy.

All nine versions are by led Zeppelin, so therefore rating one rates all nine and the criteria groups them all together, thereby I now no longer have a smart playlist of 5 star songs, I have a playlist where every 5th or 6th song is Whole Lotta Love.

Like I said, whether by design or a bug, I think it’s worth a second look by the devs.

Thanks

EB

2 Likes
4

Chiming in, I have a similar setup regarding playlists and song ratings and this almost makes Plex Music unusable for me by filling up smart playlists with live versions, alternate takes, demos etc. of favorite songs that in many cases don’t necessarily have specific alternate names.

5

I think this is something that definitely needs to be on the devs team “to do” list in a way of correcting it, or at the very least an reason why its programed like this …

My understanding from using PLEX for many years is that people who love it (like myself) do not necessarily in most cases have small music libraries, thereby the chance that they have multiple versions of the same song (at the very least a studio album versions and a greatest hits version) are the norm rather then the exception.

Thanks

EB

6

I took a few minutes to try and make a smart playlist that would solve / work around this situation till the devs had a chance to address it (which they probly should and hopefully seriously consider).

In my attempt at creating the temporary work around, I discovered and came up with some additional things that seem to further corrupt the ratings system and the results that affect the library in many different key areas.

I discovered that when a song plays, the system marks not only that song as played, but any VERSIONS of the song you have in your library as played. Studio versions, versions that appear on collections and in some cases, live versions of the song. This throws the track played category (and many others) way out of sync.

As an example, If I happen to buy an album of greatest hits from an artist. It should come up as no tracks played, but what actually happens when I add that album to my library, if I have the studio version of a song that also appears on that newly added greatest hits album, it marks those songs on that album that I have on other entires as played. This is false result.

I discovered the system marked the just added album as having songs that played 9 months ago, because the studio versions of those same songs in my library had indeed played 9 months ago.

If I made a smart playlist and the criteria was to play songs that were added to my library within the last 48 hours that had zero plays, almost nothing would play in that smart playlist off the greatest hits album I had just added. Because the systems thinks that those, or versions of those songs had already played. This also effects remastered versions of albums as well.

As you could imagine with a song, and every version of it being marked as played, as well as being assigned the same rating, this throws the entire system off when it comes to things like charts, most played songs, albums, artists and just about every other criteria that one might use to construct a smart playlist.

At the end of the day this does several things, all of them degrading the database information that formulates the listening experience, tracking of information about the library and renders the smart playlist, charts, and any other criteria information within the database and its correct usage / result.

If your using PLEX Amp 3.0 (one if the finest pieces of software for the PMS I might add … I absolutely love it), which is based on creating custom mixes, and as well as various criteria for charts, what and how you listen to your music. As time goes on and the server starts to keep track of all the songs statistics, 6 months or a year down the road, it becomes quite a task to unravel and separate out all the information in the databases.

I cant quite remember when or if this was ever changed, or perhaps it was this way from the start, and the more media I add to my library as well as the more I listen to my library it becomes a forefront issue?

Like I said, perhaps this is something that the devs and programmers should make give a look at.

Thanks for your time and also for your continued efforts

EB

7

It looks like Plex is using MusicBrainz to determine if a track is identical to another track.

If the songs are indeed identical, Plex applys the rating and play count to all of the identical tracks. Different recordings/versions with the same or similar names should be left alone.

I played:

Bananarama - I Can’t Help It (7" Version) (3:33)

and aside from increasing it’s play count, it also increased the play count of the album version, simply titled:

Bananarama - I Can’t Help It (3:28)

but it did not increase the play count of:

Bananarama - I Can’t Help It (U.S. Seven-Inch Mix) (3:39)

or any of the other five versions of the song I have (live, remixes, etc).

I found this very curious since the first two tracks were named differently and therefore were apparently different versions.

I opened all three tracks in Audacity (a sound editing app) and compared them. It turns out the album version and the (7" Version) are identical. The (U.S. Seven-Inch Mix) has a different outro.

This is something I wasn’t even aware of, but Plex figured it out for me.

It should be noted that I have a few copies of the album version on different compilations with slightly different lengths (3:28, 3:32, 3:30) and they all had their play counts increased, because they are indeed all the same track. The (U.S. Seven-Inch Mix) is roughly the same time at 3:39 but is a different edit so Plex did not increase it’s play count.

Then I went to MusicBrainz and found these pages, which confirm the album version and the (7" Version) are identical recordings but the (U.S. Seven-Inch Mix) is different:

I Can’t Help It ~ Recording by Bananarama

I Can’t Help It (U.S. seven‐inch mix) ~ Recording by Bananarama

Whether or not Plex should update play counts and ratings for multiple copies of the identical recording is up for debate, but for me, this is a feature and desired behavior.

@ebstone I discovered that when a song plays, the system marks not only that song as played, but any VERSIONS of the song you have in your library as played. Studio versions, versions that appear on collections and in some cases, live versions of the song. This throws the track played category (and many others) way out of sync.

It shouldn’t. If you have tracks with the same or similar name, but that are clearly different versions (remastered, extended, live, etc), and Plex thinks the tracks are identical, that means a user at MusicBrainz entered the wrong information about those tracks. There’s nothing you can do on the Plex side, but you can update the recording at MusicBrainz and it should filter on down to Plex.

Beware, MusicBrainz is detail oriented and not exactly user friendly, and it’s easy to make mistakes that affect everyone. If you are going to make changes, I’d poke around and try to learn the system first. Or you could ask someone in the forums there to help you fix specific recordings.

8

It goes by whatever the matched guid is.

All items, including movies/tv, that are matched the same identifiers, will have the same ratings/play count etc.

It’s the same issue that causes multiple versions of the same video to show up on continue watching (ie 4K + 1080).

9

And it’s matching the GUIDs to MusicBrainz, right? So changing the data there will change it in Plex?

10

Specifically, it seems to be going on the MB Work ID. In the case I tried, “All or Nothing at All,” both the Krall and Jarreau versions have the same MB Work ID. Artist, Recording, Release Artist, Release Group, and Track MBIDs are all different.

11

@ tikilab Thanks so much for the post.

It helps explain a lot, if indeed PLEX is using that system on its ratings, matching etc.

Do you think it should be different though? Do you agree that it throws the entire system off when it comes to creating playlists, especially smart playlists?

Perhaps what could be considered is that the devs write a script or adjustment to NOT have duplicates to appear in a created playlist, especially smart playlists. This would take care of at least that end of it (not hearing the same song multiple times in a playlist), but the charts, most played etc would still be affected by this?

I also thought that if a song is played and 3 versions of that song are marked as such, with a rating, does the database / system think that song has been played 3 times within the library?

Thanks again so much for the post and info

EB

12

This isn’t necessarily true. Musicbrainz, from what I can discern in their documentation, considers a “work” to be an original composition, distinct from performances of that work by other artists.

So, in the case of the example I mentioned, “All or Nothing at All,” the Jarreau and Krall performances are different recordings, but the same work. The MBIDs bear this out. In that case, the MB Work ID is the same for both tracks, but all other MBIDs are different.

I checked another example, a track by Enya that appears on an original album and a “Best Of” album. The tracks have identical Artist, Release, and Work MBIDs.

So the problem we’re seeing does not appear to be caused by bad data at MB.
If Plex is going on the MB Work ID when applying our ratings, that would explain what we are seeing. Perhaps it would be better if Plex used the MB Recording ID instead.

13

@beckfield If Plex is going on the MB Work ID when applying our ratings, that would explain what we are seeing. Perhaps it would be better if Plex used the MB Recording ID instead.

If they are using the Work ID, then that would be bad. But I conducted another experiment and it doesn’t seem like it.

Once again I played the song “I Can’t Help It”. Here’s the work page:

I Can’t Help It

I played the album version simply titled “I Can’t Help It”. The play count went up for all copies of that recording in my library.

The play counts for “I Can’t Help It (The Hammond Version Excursion)” and “I Can’t Help It (Extended Club Mix)” did not go up. If Plex was using Work ID instead of Recording ID, they should have increased as well because they are all listed under that Work ID.

I then played “I Can’t Help It (The Hammond Version Excursion)”. Only the play count for that recording went up. So in this case it’s working correctly.

The system isn’t perfect though. Tried it with the work Please, Please, Please, Let Me Get What I Want as well. The Smiths version, two Dream Academy versions, and a Mr. Kitty version. The Smiths and Mr. Kitty increased play count only on their specific recordings as expected (including multiple copies of the Smiths track). Dream Academy has two versions. The instrumental worked fine, but the vocal version didn’t increase the play count for multiple copies of the identical recording. That’s because MusicBrainz has them listed as three different recordings:

Please Please Please Let Me Get What I Want ~ Recording by The Dream Academy
Please, Please, Please Let Me Get What I Want ~ Recording by The Dream Academy
Please Please Please Let Me Get What I Want ~ Recording by The Dream Academy

MusicBrainz is like Wikipedia in that changes are made by the public, so the reason why there are three pages is probably because at least three different people have added this recording to their database, and they created new recordings every time. MusicBrainz wants you to own the media you are adding to the database, so if someone is adding a CD with this song, but doesn’t own this or that compilation with the same song, they can’t verify that the tracks are indeed identical. So they add it as a new recording.

So presumably someone will eventually stumble across the fact that there’s three different pages listing four different Recording IDs and they would verify the tracks are identical and merge them. (In fact, I will probably do that, after a while so whoever’s reading this at the time of posting has time to see the individual pages.)

I haven’t checked, but the problem with your Krall and Jarreau versions may be that someone mistakenly merged those two recordings together, which then lists them under the same Recording ID.

So bottom line is the data at MusicBrainz isn’t perfect, it needs to be fixed for countless tracks, and Plex is using it as-is so it will always have some mistakes. The mistakes can be fixed, but it’s going to take a lot of delicate work that most Plex users won’t be interested in.

@ebstone Do you think it should be different though?

I don’t know if treating identical tracks as one entity should be different, but I do think Plex should explain it in detail to users. It’s certainly unexpected.

@ebstone Do you agree that it throws the entire system off when it comes to creating playlists, especially smart playlists?

Absolutely. It just shows that the left hand doesn’t know what the right hand is doing, code-wise. If the library is going to “intelligently” treat these recordings as the same entity, the playlists should too.

@ebstone I also thought that if a song is played and 3 versions of that song are marked as such, with a rating, does the database / system think that song has been played 3 times within the library?

I think Plex should consider it played once, since it is grouping all the duplicates together. It might just be a matter of semantics though, because it knows there are three copies and it has to increase the play count on each one, but consider it one copy…

1 Like
14

@tikilab Thank you for your efforts in your post!

It helps and explains what we are working with, and certainly what is to be expected if no changes are implemented, which, … in a perfect world, a dev team would review what is contained herein, and at least come up with some alternative, or fix to the issue / situation.

In todays code / software world, things are written with API’s and wireframe libraries so that 3rd parties, (in this case PLEX) could write additional code to correct what is off or “customize” a function (or feature) within the software. Open GPL allows this and its made for that specific reason.

Im not suggesting the system they are using be canned. Im suggesting a piece of code be written in the algorithm on the PLEX side of things when creating / curating the playlist in the cue to eliminate duplicate songs.

Im going to go out on a limb here and venture to guess that when someone is listening to a playlist they don’t want to hear 3, 4, 5 different times the same song in the playlist because its rated once and marked on 4 tracks with the same ID.

This is NOT a small problem within the PLEX ecosystem. If the way we listen and want our music experience through the PLEX platform is based on things like ratings, times listened to, when they songs are listened to and by what artist were listened to the most, or are most popular in a particular period of time. Not having this correctly documented in the database throws the the core fundamental data needs to be correct out of balance.

If not corrected, it negates the whole purpose fo the system and when you look at a listing in your application for “the most played artist in June” that result is flawed because of incorrect data in the database from which said data is derived.

Simply put, the way the system is currently programmed, it renders the smart playlist function obsolete.

Believe me, no one more than me sympathizes with the devs and programmers. They are getting 10 different things till Sunday to fix and adjust. Its an endless cat and mouse deed to try and cover all bases to make the user base happy.

BUT, here is the difference. Many times programmers and devs cant fix why a certain function doesn’t work on a particular Windows machine when it works for 99% of the user base. But fixing or writing additional code so a database gathers and reports correct information, that’s a core function of something like PLEX.

Im hoping the devs will start looking into this and come to the community with wanting information and organized conversation about what the best solution is for the rating system moving forward.

Thanks again @tikilab

EB

1 Like
15

The way I’ve overcome is to create a mood called “SMARTRATING-IGNORE” and then add an extra step to the Smart Playlist criteria “Track Rating” is no “SMARTRATING-IGNORE”. This seems to be set at the physical file level and so avoids the issues described above

Thanks

16

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.