All We Need Is Ratio Ga-Ga

ListenersDesc.png
Figure 01: My artist-related data from Last.fm sorted (descending) by Listeners.

I find it fascinating when you think numbers seem to say one thing but beg a question that’s answered with further investigation. Often, statistics are delivered in the simplest way, out of context, and presented in a way to support what the presenter wants to say, not what the data says.

For example, in the last election cycle, a list of questions was being passed around from one group of pollsters to the other. I know this because the first time I heard them, I complained that they were poorly written and their data would be meaningless as a result. Several weeks later, another group called me and when I told the young lady I’d like to re-state what I told the earlier pollster, she said it was impossible I’d heard them before because they’d just started calling people that day. I have to paraphrase from memory and, because I’m naturally honest, I won’t be able to recall it as horrible as it actually was … it was something like, “Do you think Obamacare should be repealed?” or “What are your feelings about Obamacare?” … both with equally bad Likert Scale answers. I told the pollsters it was obvious their poll was written so that no matter who they called, their “data” would say most people hated the Affordable Care Act (also notice they called it “Obamacare”). The truth is nobody likes it but there’s no way whoever presented the data was going to mention how many people thought it was “not enough” instead of “too much” — and they didn’t bother asking.

Anyway … on the other hand, sometimes data is good but could be great in the hands of a super genius data scientist like myself.

I collect data from the last.fm API — Listeners and Playcounts for artists, albums, and tracks. While the data is yummy straight from the oven, there’s a lot of mixing and kneading I can do before baking it that makes it yummier.

Figure 01 shows Radiohead has over 7 million more followers than the artist in second-place, Queen. The difference between 2nd and 3rd place (The Cure) is even bigger with approximately 11 million less. First of all, I can’t believe Radiohead is so popular. Second of all, when I first saw their number of listeners, I immediately asked myself if that resulted in an equivalent landslide of plays.

Maybe it’s cool to like Radiohead so they have a lot of listeners, but nobody actually listens to their music. You know, the guy with vinyl editions of their albums on his wall but if he heard a Radiohead song on the, you know, radio, he wouldn’t recognize it because he’s never really listened to them.

playcountDESCmovement.png
Figure 02: My artist-related data sorted (descending) by Playcount.

The top six and the bottom six stayed exactly the same as did Meat Loaf and the Amboy Dukes. The rest didn’t change much but the big winner in the set of artists I track was Saxon, gaining 4 slots (which, because these artists are merely the union of bands I’m curious about and the Rock and Roll Hall of Fame Class of 2019 still doesn’t tell you much beyond, “Huh, that’s interesting” if even that).

But wait, there’s more.

Playcount might be equally meaningless if — and this is the outcome I was hoping for — the listeners and playcounts were really close meaning a half-billion people tried the  casserole but nobody went back for seconds.

I can already tell that Radiohead fans listen to a motherload of Radiohead music because, as it turns out, their playcount is almost a half-billion. So I wrote a little Python script to find the ratios for me which I could then use for sorting.

ratioDescMovement.png
Figure 03: My artist-related data sorted (desc) by Ratio.

The top five still hold their places. Ozzy drops two places but that’s nowhere near as interesting as Saxon moving up 12 slots, Anvil … freaking Anvil … moving up 11 … Journey losing ten places and … and … Ronnie & the Prophets gaining nine because who can get enough of Dio singing doo-wop?

Let’s talk about Evil Stig (who gained +11) because I have some algorithms-in-progress to take care of this–and other–artists.

listenersCombinedHighlighted.png
Figure 04: Merging numbers for “the same” artists

Once the app is all done, it’ll auto-magically put artists together like Dio, Ronnie & the Prophets, Ronnie & the Redcaps, Heaven & Hell, Elf, and the Electric Elves because Ronnie James Dio (actually, there are some songs for that artist specifically as well) is in all of them. Why not also add Black Sabbath and Rainbow you ask? Because he was only in those bands for some of their albums. The algorithm(s) part of the solution is relatively easy compared to the tedious, manual table-building and table-populating part. Ugh.

RJD (blue overlay in Figure 04) would move up at least a couple slots with his 6-8 bands and Joan Jett would move up at least six with her three.

Advertisements

How Do Rock and Roll Hall of Fame Nominations Affect Popularity On Spotify?

Shout out to: Malcolm MacLean‘s book, Data Driven Documents D3.js Tips and Tricks v4 which made these really super easy.

See, this here is the endgame — the kind of thing I learned D3 and coding so that I could create. This is what makes programming fun. The following are based on data I collect using the Spotify API.

Here are the inductees in the class of 2019 for the Rock and Roll Hall of Fame.

2019-inductees.png
I was in a hurry. I’ll tweak the math for legend placement later.

Here are those who were nominated but not inducted.

2019-nominees.png
I was in a hurry. I’ll put the names in a container later. Maybe.

Nominations were released October 9. You can tell I wasn’t tracking a few of these artists (John Prine, Todd Rundgren, Roxy Music, Stevie Nicks, Janet Jackson) until that date because their lines don’t start until then. There seems to be a slight gain for each (maybe) but that’s total “correlation not causation” because these are Spotify’s very relative and ever-fluctuating “popularity” scores not playcounts.

Suppose the Beatles are higher than all these artists and John Lennon says something stupid like, say, “We’re more popular than Jesus” — the Beatles popularity goes down and since it’s a zero-sum game with Spotify Popularity, everyone who doesn’t put their foot in their mouth gets a bump.

Which, looking at the above graphs, begs the question, “What happened at the end of June to make everyone crash and in mid-July to cause everyone’s climb back up?”

Another, more intriguing point, is early April when the MC5, Radiohead, and the Cure all spiked up and right back down.

Spotify doesn’t answer any of my questions so we’ll never know.

Even if nominations don’t affect anything significantly, we can watch and see if the induction has any significant effect.

Rock and Roll Hall of Fame 2018 Nominees

The latest nominees were announced in the last couple of days so I was in a sudden rush today to fix a bunch of stuff so I could get them into my database. Hence, I used the old, icky method. I’m not bummed out about that, though, because I did get a ton of stuff fixed that had become broken over time — including changing from PHP session variables to PHP cookies. I also fixed all the column-sorting links and related page-building functions that were jacked up. So, I can take screenshots of some the latest top tracks and albums for this years nominees.

Spotify, because they hate people, doesn’t make the number of plays available through their API — also not available in their web player. “Popularity” is relative to other artists, albums, and tracks on Spotify so the number may go up and down from day to day. So, from what I can guess, an artist’s top albums or tracks may not really be the top tracks or albums by that artist — because the numbers are calculated relative to all other albums and tracks — not like you’re ranking an artist’s songs by number of plays.

Yes, it’s stupid. It’s not just you.

Even more stupid is Last.fm‘s API which totally got my hopes up because you can access the number of plays! But … not the identifiers for the artists, tracks, or albums.

Because they’re stupid. That’s why.

But, in the words of Count Rogin, “Let’s work with what we have, shall we?”

Janet Jackson

janetTracks.png

janetAlbums

WordPress is ridiculously stupid. Depending on the browser and OS you’re using, the tools change. Sometimes you can control how large an image is when it’s placed. Sometimes, you can also make it so a user/reader can click it for a larger version. That seems to be increasingly rare so I apologize for how difficult to read some of these charts are.

John Prine

prineTracks.png

Each column can be sorted in the app by clicking on “Album Name”, “Track”, or “Track Popularity”.

prineAlbums.png

Again, I apologize. When an artist has a ton of albums, it makes for lots of horizontal scrolling so, at some point, I’ll make the bars listed vertically so it doesn’t suck even on a desktop.

Roxy Music

roxyTracks.png

roxyAlbums.png

Eventually, I’ll have the years appear beneath the albums. The albums, by the way, are ordered chronologically by release date.

Stevie Nicks

nicksTracks.png

I included more tracks by Nicks because I thought it was interesting that “Rooms On Fire” made it into the top five while “Talk To Me”, the other single from Rock A Little was at #23.

nicksAlbums.png

Todd Rundgren

rundgrenTracks.png

I grabbed the top 11 tracks because, again, I was rather amused by the fact that four of the top 11 were all “Bang on the Drum All Day.”

I am currently working on an algorithm that would somehow combine all the scores for the same song across albums (including remasters and compilations but not live versions). Same for albums … I’d like a score that took into account several versions of the same album.

At some point, I realized I could take screenshots of the app’s “list” view for albums. In list view, I can sort them, just like the list view for tracks.

rundgrenAlbums.png

Def Leppard

defTracks.png

I grabbed so many more of theirs because, well, I cared more.

defAlbums.png

See how there’s two versions of so many of their albums? I find that irritating. It lowers the tide for all of the albums and tracks in question.

Devo

devoTracks2.png

More rows of Devo because, again, I cared more.

devoAlbums.png

 

Kraftwerk

kraftTracks.png

kraftwerkAlbums.png

I like it when there are fewer albums. The screenshots are prettier.

LL Cool J

coolJtracks.png

coolJalbums.png

MC5

mc5tracks.png

Not a big surprise there for the #1 track.

mc5albums.png

Radiohead

radioTracks.png

radioHeadalbums.png

Rage Against the Machine

rageTracks.png

rageAlbums.png

Rufus

I know for a fact I didn’t add Rufus to my database until they were nominated last year so this line graph represents their popularity over the last year. You’re welcome.

rufusLine.png

rufusTracks.png

rufusAlbums.png

The Cure

cureTracks.png

cureAlbums2.png

The Zombies

The Zombies are nominated again which makes me happy. I think they stand a better chance of getting in this year.

zombiesTracks.png

No surprise in that list. I think what might hurt them is many people might be like me and can only name those top two tracks.

zombiesAlbums2.png

I love that Still Got That Hunger is so high on their top albums list. It deserves it. Good on ya’, guys!