Dear all,
Even after the latest update (build 200) my playcount of most played albums is still all over the place, both Tidal and own albums. Has something changed regarding calculation or am I missing something?
Thx
They should be sorted by the total amount of time spent playing that album on this profile, measured in seconds.
The time-based calculation is new in 1.3. The old algorithm looked at play count, which falsely inflated the ranking of releases made up of tons of short tracks. One listen to Bartok’s Mikrokosmos (72 minutes, 96 tracks on my recording) was receiving 19 times as much weight as one listen to Beethoven’s 9th (5 tracks, 67 minutes), when in fact they both represent approximately the same amount of investment into the material.
For people with broad collections, this tended to inflate the rankings for genres with shorter tracks (3-minute pop songs) at the expense of genres that lean towards longer forms.
We received a lot of complaints about this, so we re-designed the ranking algorithm for albums in 1.3. We went through many iterations on this in alpha before landing on this solution. It wasn’t done lightly or carelessly.
1 Like
Thx for your quick reply Brian.
Clear!
Brgds
I think there is still some room for tweaking. Maybe an interaction between time played and counts.
At the moment, most of the top listened to albums for me are operas which I listened to months ago. When I listen to operas, I listen to the whole thing at once.
So, playing an Opera once with its 3 to 5 hour play time nearly permanently cements it at the top of the list. While, albums I’ve played songs from 10 or more times don’t make it because the Ramones wrote 2 minute songs. 
One of our iterations during alpha tried to measure album plays: (seconds played/duration of album). It made things worse…
My most played album has close to 50 hours of play time. 5hrs of play on one opera would earn a ranking between 50 and 60.
I’m an extreme case–almost 30,000 plays on my profile (my first play is on May 11, 2015–the night before we launched, so I had a few hours of head start…).
The longer people use the software, the more they will start seeing behavior like mine, and the less individual plays of any album can make a big difference.
Perhaps there could be a way of focusing by time frames. For example, Last Week, Last Month, Last Quarter, Last Year, Ever", That way if I focused on Last Month, the time count would line up perfectly to my current listening habits; instead of thinking why the heck is “das gotterdammerung” my number two album, I’ve only listened to it once in the past 9 months. LOL.
1 Like
I like that idea… not sure if you can do that now with using a focus criteria?
That’s interesting…no you can’t do that right now. Focusing on “played in the last week” is totally independent from the sort order, which always looks at “all time” plays.
Hi @brian
Have you seen my posts on this “Most Played” sort option that I posted last week in Alpha??
As you can see, this doesn’t correspond with your aims above…and many of the points that applied to then B197 still apply today in B199
https://community.roonlabs.com/t/most-played-how-is-it-calculated/20423/3
@brian there has been a regression in build 209 as I now have more played albums than are actually in my library. I first noticed this the other day with the iPad update. I quit RoonServer and everything went back as normal, but now the issue has come back.
It’s not a regression–we’ve heard reports of this for a long time, and have seen it a few times, but we have no clue how to reproduce on demand, so we haven’t been able to investigate/fix it properly.
Okay, I just figured it was fixed because I haven’t had it come back the last 3-4 builds but has happened twice on build 209.
@brian I just thought of something. I haven’t been using my HQP until build 208. Don’t know if that could be a factor but figured I’d mention it.
I’m about five years late to the party, but the current algorithm for calculating most played albums generates some odd results. For example, my most played album is 10 Years Solo Live (Brad Mehdlau). I’ve played it about seven times–all five hours and six minutes. My second place album is Diamond Mine (King Creosote). I’ve played this album 30 times but, as it’s only 32 minutes long, it’s being relegated to second place. In this instance I would have thought it obvious that Diamond Mine is more “most played” than 10 Years Solo.
I can see the rationale for not using the total tracks play count–that makes complete sense if you have an album with lots of short tracks–but skewing the calculation so heavily towards total play duration causes a similar error.
Would it be possible to base the album play count on the average track play count for each album? A modal or median count would work well, i.e. prevent the algorithm being skewed by a large play count for just one or two tracks.
The issue was already noted by me upthread back then.
And at this point, you are asking a Feature Request and should post your suggestion in that section.
1 Like
Sorry, running around getting ready for the holidays and was on my phone earlier, so, not as conducive to long text strings. In a nutshell, as you see from the post I quoted, I pretty much agree that the current compromise has it’s issues.
I still think that straight counts, or runtime, are going to have their own different corner case issues. You can, of course, remove some of the offending plays from History to tweak the stats to your liking.
1 Like
This topic was automatically closed after 21 hours. New replies are no longer allowed.