About Song Metrics

This whole thing started just because I really like listening to The Head and The Heart. I have a playlist dedicated to crying (because who likes crying in silence? No, crying with background music is far superior) and many of their songs feature on the playlist. If anyone cares, this is my all-time favourite performance of theirs. So back when they were releasing their latest album ‘Every Shade of Blue’, some people started commenting about how their ‘sound’ has changed, how they wanted music more like their earlier albums, etc. To be fair, I think that’s the case with a lot of artistes — some people will always like the earlier stuff more than the newer releases and vice versa.

I don’t know if this counts as an unpopular opinion but I think people should be allowed to not like newer songs that artistes they like put out without being slammed or I don’t know… being called ‘fake fans’, the same way fans shouldn’t demand artistes not change and accuse them of being sellouts or something. Ben Howard’s old songs, for example, sound very different from his later releases and while I prefer the old songs (’Keep Your Head Up, ‘The Fear’, etc.), it doesn’t mean I take his subsequent music as some kind of personal betrayal.

I mean I get it, with a lot of these kinds of music, they mean something to the listener so I guess if the artist changes where they are but you’re still back where you always were or where you want to still be, then it kind of feels like being left behind. The songs also mean something to the artistes though, so if they’re not where they used to be, it’s not their fault either. I’m not sure I’m making much sense, I’m rambling at this point. To be honest, part of this stems from a Reddit post some years back that was talking about how at concerts, people will yell at Ben Howard to perform his old songs even though he has distanced himself from the first album. Again, I get that people like the first album songs but it is also shitty to talk over and disregard performances of the other songs, especially at their concert (this bit shouldn’t be an unpopular opinion).

Anyway, I digress. Going back to The Head and The Heart, I do like many of their songs, old and new, but I admit that the newer songs haven’t been included in my crying playlist (just not the right vibes for crying). So just for kicks, I thought I’d look at the metrics of the tracks from the different album to see how different they are (intuitively, I do believe that the lyric development across albums would play a much bigger role in this story but I’m going to leave that for another day). There isn’t much analysis here, this is purely descriptive and there are a couple of interactive plots for you to see specific track metrics if you’d like.

I do want to make a final point which is that having done all this, at the end of the day, I don’t think music can/should be distilled quantitatively like this (full disclosure: I don’t know much about music. I don’t play any instruments, got kicked out of multiple beginner classes when I was trying to learn though that might have more to do with attitude than anything else, and I failed all the compulsory music classes we used to have to take in school). All these properties contribute to making you feel some kind of way sure, but I guess some days I don’t want to think too much about these characteristics. Most days I just want to cry, dance, feel alive, pretend I’m in a movie living a different life, use it to be passive-aggressive to people (language NSFW), plan my wedding, or my funeral. To just feel and not think about why (that’s the whole point of the head versus (and?) the heart, no?)

Alright, here we go.

Overview of albums and metrics

For the purpose of this, we’ll only use the 5 studio albums (i.e. no live versions etc). In chronological order of release, they are ‘The Head and The Heart’, ‘Let’s Be Still’, ‘Signs of Light’, ‘Living Mirage’, and ‘Every Shade of Blue’. As for the metrics, we’ll use ‘valence’, ‘danceability’, ‘tempo’, and ‘energy’ from Spotify’s API. Do I really know what the metrics are about? No (recall: failing every music class). Will that stop me from continuing with this? No. Is that right? Meh. I’ll talk a little bit about ‘valence’ anyway, since that seemed the least intuitive to understand.

According to Spotify, valence is:

A measure from 0.0 to 1.0 describing the musical positiveness conveyed by a track. Tracks with high valence sound more positive (e.g. happy, cheerful, euphoric), while tracks with low valence sound more negative (e.g. sad, depressed, angry).

This shouldn’t be confused for actual ‘positive-ness’ of the track. For example, the valence of Hollywood Undead’s Bullet is 0.889 but this isn’t what the general public considers a ‘positive, uplifting’ song (TW: self-harm/suicide). Side note: having said that, people have also mentioned that this song helped them through difficult times, in part because it made them feel less alone so there’s something to be said about what’s actually ‘positive’. Again, the point is that valence refers to how positive it sounds musically (I don’t believe it takes into consideration lyrical content).

It’s still unclear, however, what ‘musical positiveness’ actually means. What does it take for a track to sound more positive or negative? I would have thought it would be somthing like how upbeat the song is but it’s supposedly not about the tempo or energy of the song, since those are separate metrics. As for the other metrics —

Danceability:

describes how suitable a track is for dancing based on a combination of musical elements including tempo, rhythm stability, beat strength, and overall regularity. A value of 0.0 is least danceable and 1.0 is most danceable.

Tempo:

The overall estimated tempo of a track in beats per minute (BPM). In musical terminology, tempo is the speed or pace of a given piece and derives directly from the average beat duration.

and Energy:

a measure from 0.0 to 1.0 and represents a perceptual measure of intensity and activity. Typically, energetic tracks feel fast, loud, and noisy. For example, death metal has high energy, while a Bach prelude scores low on the scale. Perceptual features contributing to this attribute include dynamic range, perceived loudness, timbre, onset rate, and general entropy.

The chart below is an overview of the albums, the metrics and the various relationships present.

The most obvious thing to note is that the number of tracks in album increases from the earlier to latest releases (the people said ‘we want more’ and they were like ‘okay’). Given the definition of danceability as being, in part, based on tempo, I expected that the two would be somewhat correlated but that doesn’t seem to be the case here. Instead, valence is positively correlated to both danceability and energy (I suppose it might be possible to model valence as a function of the others to figure out what valence is…) ‘Signs of Light’ in particular has the strongest positive correlation for both (0.909 for danceability, 0.748 for energy). With the exception of valence, the interquartile range is also the smallest in ‘Signs of Light’ compared to the other albums, suggesting less variability but we can take a closer look at the tracks in each album anyway.

Valence

If you hover your cursor over each point, you should be able to see the valence for each track (of course I had to make the graph every shade of blue. Okay well not every, just five). I greatly enjoy the fact that across all the albums, the song with the highest valence is ‘Running Through Hell’ while the lowest valence song is ‘Signs of Light’. I suppose you need a bit of pep in your step while you run through hell.

Danceability

I also know nothing about dancing (does this take into consideration specific kinds of dance? Like contemporary versus hip hop? Jazz? Can you break dance to ‘Paradigm’? Is ‘Glory of Music II’ good for raves?) Again, I find it hilarious that on ‘Let’s Be Still’, the track with highest danceability is well… ‘Let’s Be Still’. In any case, it seems to be that if you want to dance, your chances are better with the later two albums than the earlier ones.

Tempo

This time, I included x-axis breaks at the supposedly most common tempo markings (bpm range). Make of that what you will. It seems a little like the variability decreases with the newer albums. Still funny to me that slowest tempo song across all albums is wait for it… ‘Taking My Time’. They really know what they’re doing, don’t they.

Energy

Last plot here is for energy. Like before, more of the songs in the later albums cluster near the higher end of energy while songs on the first two albums seem more spread out. ‘Hurts (but it Goes Away)’ is the highest energy on ‘Every Shade of Blue’, which is my favourite of that album but ‘Rivers and Roads’, which is my favourite of ‘The Head and the Heart’ is the lowest energy so I don’t know what that says about my taste. Probably that I’m fickle, which is not brand new information. They’re both just really comforting in their own way.

So that’s all for now. If I ever pull myself together to do it, there’s probably more to unpack in the lyrics but we’ll have to wait and see if I ever get around to doing it. It’ll probably happen when I have some other pressing deadline to meet — that’s the best time to do other random things.

Postscript

I did this over a year or so ago (in bits and pieces, I kept putting different parts of it off). Part of me did think about whether I should still post this because it’s not great in terms of storytelling or analytics. Code could probably also be a lot cleaner. I like to think I’ve learned a little more since then (just a little bit) but I think there’s value in putting rough starts out there. This took me forever in terms of just figuring out how to connect to the Spotify API and then to clean the data to get the five albums as they are (what with all the multiple versions of the same song). So I guess I’m allowed to be a little pleased that I managed it all in the end. As for having any kind of coherent message, this was never really about the trajectory of music, what it all means, or me trying to get better at coding. I just wanted to be able to talk about some songs I like. Psych.