Measuring the Unmeasurable: Ranking One’s Favorite Music, Part III

In June and September of 2021, I wrote the first two essays in a series on the evolution of the methods I use to rank my favorite tracks (a term I prefer to “songs”). These essays presented the history of the 309 individual “mix” cassettes, videocassettes and CDs I constructed between August 1981 and August 2016. Onto these media I recorded 3,381 unique tracks covering 6,335 slots; the average mix had 21 tracks, of which 11, on average, appeared for the first time. Through 2005, the last year in which I systematically ranked my favorite tracks – and thence, albums and artists – I used only data from these mixes for analysis.

Undergirding this analysis was a set of “ranking hypotheses,” which boil down to:

  1. Appearance on any mix signifies “liking” a track (relative to tracks never recorded onto a mix) – though some mixes count more.
  2. At time of recording, I liked some tracks put on a mix more than others.
  3. The more mixes include a track, the more I like it.
  4. The longer a track goes without a new appearance, the less I like it – unless I simply have not had the chance to re-record it.
  5. Too many appearances in a short period of time (past burnout) and too recent an appearance (future burnout) reduce how much I like a track.

These are reasonable hypotheses given how labor-intensive creating mixes was through 2005. Still, as I wrote in the September essay:

I will spare you the convoluted algorithm I ultimately devised, with “future placement” becoming an easy – and highly subjective – way to increase or decrease a track’s final score if it lacked “face validity,” meaning it just didn’t feel right, particularly when ranking tracks overall, or within artist or year of release. The point being that as much as I tried to use an objective algorithm to construct a valid score for each track, I was already making hard-to-defend adjustments to individual elements. My straightforward notion – adjusting presence on a mix by time-of-recording affect; passage of time; and various measures of burnout, future recording and recency – proved extremely difficult to execute. The individual measures were reasonable – the final score had reasonably high construct and content validity – but my “objective” algorithm had far more “subjectivity” than I had anticipated.

Things were about to get even more complicated, however, for the simple reason that I really like taking long hot baths.

**********

I have always enjoyed hot baths, so when two buddies and I moved to an off-campus apartment for our senior year at Yale, I took advantage of its tub to take one every night; this routine continued for eight years and two additional apartments. However, after moving into a Somerville, MA apartment with my then-girlfriend in September 1995, nighttime baths became more sporadic. When our relationship ended in February 2001, I moved into an apartment complex with an outdoor swimming pool in Philadelphia; I soon began to go for a swim after work, then take a hot bath to unwind more. With more portable technology and no roommates to bother, I began to play music while I bathed, zeroing in on St. Germain, The Smiths and, especially, Miles Davis, having purchased the shimmering Blue Miles after hearing it the previous Thanksgiving.

The following February, I moved to an apartment with a magnificent clawfoot bathtub, and I began to limit my baths to Friday nights. I do not recall when I first schlepped my CD/cassette/AM-FM “boom box” into the bathroom so I could listen to the “mellow” Side II of Stuff and Such Vol. XLVI – but that was a revelation. One year later, I moved to a much nicer apartment in King of Prussia, albeit one lacking a magnificent clawfoot bathtub. A substantial increase in annual income allowed me to purchase a more powerful desktop PC, a portable CD player and computer speakers; it was here, in December 2004, I reconstructed my Microsoft Excel workbook of mix-related data.

It was also here I began to make CD mixes, though I did not burn multiple CDs at one time until November 2003, when I created CD Stuff Vols. III-V. For the first and only time, the opening mix was “mellow;” CD Stuff Vol. III, opened with Dave Mason’s “We Just Disagree” and closed with The Blue Nile’s “Easter Parade.” This CD and the “rocking” CD Stuff Vol. X, burned in June 2004, became bath music staples. It was easy to change discs on the portable CD player, plugged into the computer speakers, skipping tracks and adjusting the volume as I saw fit. And at some point, I first brought into the bathroom with me a tumbler filled with Scotch whiskey, typically Johnny Walker Black Label, over ice – along with a very large cup of ice water – which I drank as I soaked and listened. This was how my Friday night bathing ritual was born.

I perfected the ritual after I moved back to the Boston area in the fall of 2005: I added large, glass-jarred pine-scented candles, allowing me to turn off the lights, and Village Naturals Lavender Milk Bath. Soon, I began to turn off every light in the apartment before entering the bathroom, so I could carry my tumbler of Johnny Walker Black into the bathroom, eyes closed and completely naked. The candles had already been lit, and the CD player cued up; when I opened my eyes, I saw only a dim tableau.

And if this sounds like self-medicating – it was. As we see later, I quickly began to eliminate this alcohol-soaked ritual after I began to take the antidepressant Effexor in October 2016.

**********

At first, I continued to bring pre-recorded CDs into the Waltham bathroom with me. Until one day in 2006, that is, when I began to experiment in Musicmatch Jukebox with mixes made specifically for my Friday night baths; I maintained the ritual despite pursuing a master’s degree in biostatistics full-time at Boston University School of Public Health (“BUSPH”). What I wanted was a purely melancholic feel: “zone out” music that matched work-week fatigue and undiagnosed depression. An Excel spreadsheet records the first three “experiments,” though I do not recall burning them onto CDs. It is likely I was trying to whittle down a group of tracks so they fit onto one or more 80-minute CDs: the first three “experiments” comprise 29 (total time: 142:48), 13 (58:28) and 17 (89:58) tracks, with a fair amount of overlap.

What I do recall is burning the following pair of CDs, most likely one late summer day; there is even more overlap between these two and the final two “experiments”:

Bathtub CD 1 (Summer 2006)Total time: 72:52
We Just DisagreeDave Mason
BeautifulGordon Lightfoot
Drowning ManU2
The Carnival Is OverDead Can Dance
It Was a Very Good YearFrank Sinatra
DarknessThe Police
TenderlyDuke Ellington & His Orchestra
Constant Cravingk.d. lang
Willow Weep For MeDuke Ellington & His Orchestra
Prelude To a KissDuke Ellington & His Orchestra
Last Chance on the StairwayDuran Duran
Sweet PeaMiles Davis
Blue In GreenMiles Davis
Dream All DayThe Posies
All I KnowArt Garfunkel
Stormy Side of TownStan Ridgway
  
Bathtub CD 2 (Summer 2006)Total time: 75:06
So WhatMiles Davis
Half a PersonThe Smiths
Walkin’ Home AloneStan Ridgway
With You I’m Born AgainBilly Preston
MainstreetBob Seger & the Silver Bullet Band
Black MetallicCatherine Wheel
Blue Spanish SkyChris Isaak
Beyond DoubtGene Loves Jezebel
Fly On a WindshieldGenesis
Pussywillows, Cat-TailsGordon Lightfoot
It’s A LaughHall and Oates
Don’t ChangeINXS
The Better IdeaIt’s Immaterial
Too HotKool & the Gang
There Is a Light That Never Goes OutThe Smiths
Easter ParadeThe Blue Nile

It is no coincidence the bookend tracks match those of CD Stuff, Vol. III. Note also multiple tracks by The Smiths, as well as by Gordon Lightfoot and Stan Ridgway. Meanwhile, six of these 33 tracks – three each by Davis and Duke Ellington & His Orchestra – are jazz instrumentals, continuing the Blue Miles trend. The proximity of Ellington and Davis tracks to each other – and the alphabetical ordering of artists on the second disc – suggest I was not putting much effort into sequencing. Nonetheless, this was a breakthrough mix – and not only because these 33 tracks still rank among my few hundred favorites.

For a forgotten number of lavender-scented Friday nights, I absorbed the music on these two discs, extending the ritual to at least 2½ hours. Then, on October 12, 2006, I recorded a new, single-CD bathtub mix; this is a Thursday, so I likely burned it a day in advance. None of its 14 tracks – covering just 56 minutes and 1 second – had appeared on the previous mix, suggesting some burnout. Two weeks later, I burned an 18-track mix which nearly filled the disc: 76 minutes, 22 seconds. Five tracks overlapped this time, including the opening track of the October 12 mix: “Why Go It Alone?” by Pere Ubu.

I burned new mixes on November 2, 10 and 16 (16 tracks each, ~71 minutes), then none until December 30 (16, 72:34). Between January 19 and June 29, 2007, I burned 12 more bathtub mix CDs. The 18 total mixes I had burned in less than a year contain 194 unique tracks over 331 slots. One stands out: “The Carnival Is Over” by Dead Can Dance. This iridescent track appears on 13 mixes, opening 12, including nine in a row from February 2 to May 11, 2007.

This bathtub mix-making occurred during a particularly eventful year. On Halloween night 2005, I met a woman named Nell on the now-defunct Friendster website. We dated until mid-May 2006, then separated until early October. Just after completing my degree in January 2007, I started a full-time salaried position. Later that month, Nell told me she was pregnant. We soon learned it was a partial molar pregnancy; it was terminated on Valentine’s Day. This tragedy brought us even closer togther, and I proposed on June 2; she accepted happily. We moved into a new apartment in Brookline in late August, two months after I burned the 18th bathtub mix.

For some reason, I reverted to listening to prerecorded CDs, creating only eight new bathtub mixes through the end of 2013. I actually created the first three while I bathed, using the iPod I had plugged into the computer speakers. These playlists were built around 1) The Waterboys’ Fisherman’s Blues, 2) new wave tracks from the 1980s and 3) Davis. I bought the Waterboys’ CD in New Hampshire with Nell one night in 2007 after we had recently heard much of it in the backroom of The Burren; this masterpiece is basically “our” album.

I created the next five mixes on my computer, though I did burn CDs; the iPod sufficed. These were mega-mixes of between 40 and 55 tracks; their 250 total slots comprise 167 unique tracks (with Jerry Goldsmith’s “Theme to L.A. Confidential” appearing twice on two mixes). Rather than create new mixes, I simply cued up the most recent mega-mix, sometimes skipping tracks I did not want to hear again. Even so, these baths were now regularly topping the three-hour mark; the Scotch was basically water when the final track played.

By the end of 2013, I had burned and/or constructed 26 bathtub mixes consisting of 357 unique tracks in 641 total slots; the averages of 24.7 slots per mix and 1.8 appearances per track are similar to those for the “traditional” mixes I was still constructing. Which makes it all the more embarrassing that when I began to enter data for these mixes, I created an entirely new Excel workbook, one which had slightly different track titles and artist names, and with no “crosswalk” variable to link the two “mix” datasets. I mentally separated bathtub and traditional mixes because I made the former more quickly, and I only used a “mellow” subset of tracks – though this was not strictly-speaking true.

Table 1: Bathtub Mixes Constructed by Year, 2006-20

Year# Mixes# Slots# TracksNew Unique TracksCumulative Unique Tracks
20066129989898
20071220211896194
2008-12620513497291
201321059766357
201410295209147504
201531682581463967
2016194293672081,175
2017111911581011,276
201848473301,306
20191272781,314
2020515282431,357
TOTALS1072,5011,3571,3571,357

Still, as Table 1 shows, the number of bathtub mixes increased rapidly over the next four years, during the peak of my ritual bathing. This begs the questions: why the sharp increase in 2014 – and why the sharp decrease in 2018?

**********

At some point in 2013, the PC I bought in 2003 died; its replacement was a substantial upgrade. This was also around when I began to digitize my vinyl album collection and acquire many new albums, in multiple formats, when Nell sold her mother’s house and moved her to an elderly-care facility much closer to Brookline. In fact, the number of tracks in my iTunes increased from 5,350 in July 2010 to 8,767 in March 2015.

But the way in which these tracks were organized appalled me. Which is why, sometime in 2014 (even as I was completing my epidemiology doctorate at BUSPH and wrapping up my data project manager job at Joslin Diabetes Center), I began to “clean” my iTunes using the “Song Info” option. My goal was for every track to have the correct title and artist name (using “last name first” and “BandName, The” formats); mode and year of first release (e.g., studio album, EP, single, song first written); location on first release; and correct artwork for first release.

Moreover, I wanted every track to have ≥1 plays; I was just beginning to pay close attention to number of “Plays,” even if it only started in 2013. To advance this process, 10 of the first 11 bathtub mixes I created in 2015 consisted solely of iTunes tracks with 0 plays. These mixes – which were “need to hear,” not “want to hear” – helped me to rationalize maintaining two separate mix databases; I was also daunted by the work involved in combining them. I finished cleaning my iTunes sometime in 2015 – just as I was retiring from Joslin and, as it turned out, my career as a health-related data analyst.

Meanwhile, the bathtub I used in our first Brookline apartment was downstairs, with a thick wall separating it from the nearest bedroom. I had no qualms about playing music for up to four hours every Friday night, adjusting the volume mostly as I pleased. In October 2016, though, I began taking Effexor daily, and all that extra serotonin slowly led me to drink less and less alcohol; there was no longer a need to self-medicate. Then, in July 2018, we moved to a different Brookline apartment, and the even-more-magnificent standalone bathtub was just off the bedroom I shared with Nell, an early-to-bed person. This pretty much spelled the end of bathtub mixes, at least when Nell was home.

Thus, after averaging nearly 18 bathtub mixes from 2014 to 2017, I have made only 10 since then, and none since August 29, 2020. I was still taking frequent hot baths, of course, just nearly always without the alcohol and music.

I was also constructing other non-traditional mixes. From 2011 to 2020, I constructed a total of 19 annual birthday mixes for our two children,[1] introducing them to different musical artists through tracks I thought they would like. While I burned an actual CD for their “birthday bucket” of amusing chotchkes – here is the aftermath of one such bucket – I actually played the mixes on my iPod in the car as we drove. This had the unintended effect, though, of “artificially” increasing the iTunes Plays count for these tracks – including ones I would not have put onto a mix intended solely for me.

I describe the genesis of my Thanksgiving cleanup mixes elsewhere, meanwhile, so I simply note I constructed new ones in 2014 (50 tracks), 2016 (50), 2017 (57), 2019 (50) and 2021 (37). Finally, I constructed mixes for family drives to Philadelphia in 2015 (117 total tracks) and 2017 (50); the former was actually seven distinct playlists: the traditional Drives I and II plus one consisting only of new wave/synthpop tracks from 1983, and one each consisting only of tracks by Davis, Ridgway, Bruce Springsteen and The Waterboys. Excluding the birthday CDs, that is an additional 13 “mixes.” The construction and enjoyment of every one of these mixes also increased the count of iTunes Plays for these tracks.

All told, then, I created 123 non-traditional mixes between 2006 and 2021, including the three wedding mixes; including birthday mixes increases the total to 142. Moreover, I have ONLY made non-traditional mixes since August 2016: 21 bathtub, one family travel and three Thanksgiving – and seven birthday. To the extent a mix captures a set of tracks I particularly want to hear at a given moment in time, the distinction between “traditional” and “non-traditional” was rapidly blurring.

**********

That I understood the distinctions were collapsing is reflected in the ways I began to think about how I could use bathtub mix data to “adjust” traditional track. I settled upon calculating a multiplier: 1 plus the quantity total appearances divided by 100. For example, the gorgeous “Beyond Doubt” has thus far appeared on 11 bathtub mixes, tying if for 8th with “Him” by Rupert Holmes; “The Carnival Is Over” leads with 23 appearances. Dividing 11 by 100 gives you 0.11, which added to 1 is 1.11 – the value by which I would have multiplied its traditional score. I also experimented with weighting this adjustment by 1) number of bathtub mixes made in a year and 2) total years in which I recorded the track onto a bathtub mix. This latter Bathtub Mix score later became one of four values – along with Adjusted Total Plays (explained in the next essay), Media Mix (i.e., traditional) and Other Mix (a weighted combination of appearances on wedding, family drive and Thanksgiving mixes).

This seemed like a reasonable approach through the summer of 2021, when I finally girded myself to combine data from all 432 non-birthday mixes into a single database. This took some doing, mostly because of mismatches in track title and artist name, but it was worth the effort. Table 2 summarizes these data.

Table 2: Summary of Traditional and Non-Traditional Mixes, 1981-2021

Year# Mixes# SlotsCumulative Unique Tracks
1981-2005 (Trad only)1853,9662,136
2006275102,325
2007285162,479
2008-12408622,855
2013265743,093
2014235813,223
2015509333,387
2016286293,400
2017-2021 (Non-trad only)251993,595
TOTALS4328,7703,595

A total of 3,595 tracks appeared on at least one mix of any kind between August 1981 and November 2021, with 212 (5.9%) never appearing on a traditional mix. Of the latter tracks, nearly half (105) first appeared on a mix after 2016, meaning they have no traditional score to adjust, solidifying my decision to create a single dataset. Overall, the average number of tracks per mix is 20.3, while each track now appears on an average of 2.4 mixes. These 3,595 tracks represent 36.9% of the 9,730 tracks with ≥1 iTunes Play – the complete universe of tracks to be ranked.[2]

In the next essay, I will describe how I rejected combining four measures into a track score and decided to combine only two: Adjusted Total Plays and Media Mix Score. I will also describe how I streamlined the process of combining mix appearances into the latter score, rethinking some hypotheses along the way.

Until next time…please be safe and healthy! And if you like what you read here, please consider making a small donation here. Thank you!


[1] Both are biologically female, but the younger one currently considers themselves gender non-binary.

[2] Technically, it is 9,912, but I have not yet cleaned/played 182 of them.

One thought on “Measuring the Unmeasurable: Ranking One’s Favorite Music, Part III

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s