Of Pitchfork’s Top 100 Songs of 2023

The journey begins 

Those who browse my Consumption Log—recently moved into a new home as opposed to its old higgledy-piggledy single, ever-expanding post—may notice a continued thematic omission: I post little music. This is not to say I don’t listen to music, only that I have fallen into the all to common trap of treating music not as something to be enjoyed in its own right, but as a lubricant for other activities: work, leisure, chores are all carried out with some sort of music in the background. 

This is all very well and good, but does not lend itself to the air of mindful consumption I attempt to cultivate these days. It also does not lend itself to me discovering and refining my tastes in any meaningful way. Hence a paralysis and what must seem as evasion whenever someone asks me what my favourite music is.¹Ah, to be able to confidently say my favourite music is Band X and be done with it. Although in writing that, I’m aware it’s almost an antisocial sentiment: that I want to forestall a conversation and simply provide a verbal tick box response. Of course, the idea of having a favourite anything has always struck me as deeply impossible, the same (feigned? actual?) paralysis when someone asks a parent what their favourite child is. 

In any attempt to switch my media consumption from mindless to if not mindful per se, then certainly more intentional, I decided to listen to, and comment on, Pitchfork’s 100 Best Songs of 2023. 

This turned out to be oddly timely: shortly thereafter, the announcement came that another Condé Nast property, GQ, would absorb Pitchfork. Because only ‘gentlemen’ like music, right? Truly absurd. While I was never a habitual Pitchfork reader, I appreciate and value its existence. As I’ve written before, we need more criticism, not less. The decision is sadly typical of mass media’s disdain (or discomfort) for criticism, and the tensions it raises between webs of commercial obligations.

About the experiment 

The process proved fascinating: I discovered a lot of music that but for this exercise I would not have encountered. I also discovered listening to music as an end in itself is difficult for me. To concentrate solely on music is something I found I could only do in short bursts, hence the entire process of listening to all one hundred tracks took much longer than I had expected. This was partially procedural—I listened to each song twice, once with my focus entirely on the song and then again while following along with the lyrics, sourced from Genius.²I confess I did not always have the discipline to focus entirely on a song for the first time I listened to it.³If you’ve never spent time looking at song lyrics on Genius, it is a delightful experience. Users are allowed to annotate songs, which sometimes genuinely helps one understand a reference (see the note for “Mixin’codeine up with the Phenergan”) but often turn out to be rather elementary explanations that I suspect no one needed (cf the note for the line “Heavily, we are getting lit”).

For each song, I marked whether I could identify the artist as BIPOC, gave it a rating, made a comment (a mini review, mostly) and assigned it a keyword (a reference to help me remember the song or sometimes a joke that amused me). The ratings are based on my standard Consumption Log three-point scale, although with their meanings slightly changed:

• 🙆 A song I quite enjoyed and would listen to again as part of my regular music life. 
• 🤷 A song I enjoyed, but for whatever reason, I didn’t think I would have the desire to listen to it again.⁴As I write this I’m listening to these songs to see if any were misjudged—thus far I’m happy with my original decision.
• 🙅 A song I did not like and have no intention of listening to ever again.

The Results—Music

I liked only a third of the Top 100; overall the majority of songs (66%) I was either cool on or simply didn’t like. The songs I did like were pretty all over their place in terms of their overall popularity on Spotify:⁵Spotify has a field called Popularity, which is essentially an index of 1 to 100 based on how popular a song is (at a particular point of time – I refreshed these values on 20 February 2024 to have the latest data).

I compiled playlists if you wanted to have a listen to the songs I liked, the songs I thought were okay, and the songs that I didn’t care for:

Conclusions about the list

I loved this exercise. I found a bunch of new music that I really enjoy. And it has inspired me to continue to be curious and explore music. It’s all too easy to let one’s taste ossify: the meme of an old man only liking the music from their childhood has real truth to it.

It’s fair to say Pitchfork does not map 1:1 to my tastes. This is okay though, as clearly there is enough on the list that I did like. I was surprised that there was such a small percentage of songs in languages other than English. There was only one song in an Asian language (Korean), a few songs in French and a few songs Spanish. I was happy to encounter a few songs in patois, but overall I was yearning for a little more diversity in the languages of the songs on the list. I wonder if this reflects Pitchfork’s biases or broader biases in music publishing?

Thinking about the music I did like on the list, there was not a lot that surprised me. The songs I liked seemed like songs I would like, which suggests a certain fixity may have already set in. Yet being mindful of this fixity while remaining open to new experiences is key to a happy, rich life.

I would encourage everyone to do an experiment like this. It’s a lot of fun, a great learning exercise, and a nice side-project.

You can (should?) stop reading now, or join me in my madness below…

The results—Data Madness

I was curious to see if there were any commonalities in the songs I assigned to each rating: did I have a hidden preference towards one type of song other than the others? Reader, here I fell into a rabbit hole that would threaten to derail this entire project, if not the course of my life. 

You see, like many of us, I do a bit of data work at work: I’m certainly not a trained data scientist (or even analyst), but I can do some basic analysis that often suffices for the need at hand. I don’t mean to brag, but I’m quite a dab hand at Excel. Pivot tables are my thing!  

So I started my analysis. I quickly found there was no there were no significant statistical difference between median length or explicit language of songs I liked versus songs I didn’t. I had a slight bias against non-BIPOC artists: 66% of the songs I didn’t like (i.e. 🙅 songs) were from non-BIPOC artists.⁶I acknowledge that my accuracy with noting whether an artist is BIPOC is quite imprecise, and so expect inaccuracy. The identity of those who make our media is an issue I think we ought to remain aware of, as I wrote here. If I did this again, I would attempt to flag queer artists, too.

Spotify has many fields that it uses to drive its recommendation algorithms.⁷Something which all music services have done a poor job at. Human curation remains supreme—quelle surprise.

Using the popularity field, I made a histogram for each of my three ratings to see if my tastes broadly align with those of others on the platform. At first I thought my favourites were normally-distributed, with the other two ranks were right skewed. And then, thus proving my point about my so-so data skills, I realised the second two charts had a different x-axis scale, and so when I made sure each of the three charts had a common scale, they were all largely normally distributed. 

Then, I looked at a host of internal spotify fields—each track on spotify has a host of unseen fields called things like “speechiness,” “valence”, or “danceability”—to see if there was any correlation between my tastes and these values and ran into a roadblock: I couldn’t quite figure out a way to efficiently and usefully analyse the data to expose any sort of meaningful insights. This could have been a skill issue, and/or there is simply no rhyme nor reason (at least on a statistical-level) for the songs I liked.

At any rate, I’d need more than Excel to proceed. So I downloaded R, a programming language for statistical analysis and data visualisation. I was immediately out of my depth. While I had done a day’s introductory training to R a few years ago, I remembered almost nothing. And like many super powerful programs, it is far from intuitive.

I quickly came to realise my problem was twofold:

1) I didn’t have the statistics background to be clear what I wanted to do (or even should do); and

2) For the ideas I had, I wasn’t sure at all how to execute them.

Unlike Excel (excluding VBA), R functions as a programming interface, requiring users to enter text commands into a command prompt. For the initial part of my analysis, I made the mistake of using R itself, instead of the slightly more user-friendly R Studio.

Eventually, after many beginner mistakes, I imported my data and made one of the world’s most confusing scatterplot matrix: 

It was at this stage I realised I would need to walk before I ran. The hidden insights of this data would not jump out to me as I had naively hoped. I would have to learn R from scratch. An online friend pointed me to a helpful online textbook, and so over the next few weeks I’m going to work my way through it and understand whether there is a method to my musical madness tastes.

The Table

If you have read this far, I applaud your commitment and perseverance! Below is the table giving my full observations on all 100 songs (and here is the google sheets version with more metadata available, should you wish to do my homework for me 😳👉👈)

Notes

• 1
  Ah, to be able to confidently say my favourite music is Band X and be done with it. Although in writing that, I’m aware it’s almost an antisocial sentiment: that I want to forestall a conversation and simply provide a verbal tick box response.
• 2
  I confess I did not always have the discipline to focus entirely on a song for the first time I listened to it.
• 3
  If you’ve never spent time looking at song lyrics on Genius, it is a delightful experience. Users are allowed to annotate songs, which sometimes genuinely helps one understand a reference (see the note for “Mixin’codeine up with the Phenergan”) but often turn out to be rather elementary explanations that I suspect no one needed (cf the note for the line “Heavily, we are getting lit”).
• 4
  As I write this I’m listening to these songs to see if any were misjudged—thus far I’m happy with my original decision.
• 5
  Spotify has a field called Popularity, which is essentially an index of 1 to 100 based on how popular a song is (at a particular point of time – I refreshed these values on 20 February 2024 to have the latest data).
• 6
  I acknowledge that my accuracy with noting whether an artist is BIPOC is quite imprecise, and so expect inaccuracy. The identity of those who make our media is an issue I think we ought to remain aware of, as I wrote here. If I did this again, I would attempt to flag queer artists, too.
• 7
  Something which all music services have done a poor job at. Human curation remains supreme—quelle surprise.