Skip to content

← The archive

Five Years, 116,000-odd Scrobbles, and an Entirely Reasonable Number of Wolves

A snapshot of my music scrobble history, I guess.

Every track I play ends up in fm.teal.feed.play, a collection on my own AT Protocol repository at eurosky.social: one record per play, with the track, the artist, the album, a timestamp and, lately, a duration. No walled garden holds it, no algorithm buries it — it's a plain, queryable log that I own. This week I sat down, fetched the whole thing — 116,000-odd records and counting — and asked the obvious question: what do I actually listen to?

The answer is a bit more complicated than "I scrobble everything", and more interesting. Part of this history is live, part of it is rescued, and the seam between the two is visible in the data. But first, the numbers.

The headline numbers

That listening-time figure needs a caveat: only about 23,000 records carry durations, all from my Apple Music era, so the total extrapolates the observed average of about 3.4 minutes a play across everything else. And since the collection keeps growing by the hour, I've rounded anything that can still grow — exact figures would be out of date before this post finished uploading. Numbers from closed years stay exact; those will never change. The plays sort into six eras — one a year, each with its own soundtrack.

Scrobbles per year, bar length proportional to plays:

2021 | █ 1,515
2022 | ███████ 10,276
2023 | █████████████ 18,195
2024 | ██████████ 14,178
2025 | ██████████████████████████████ 42,722
2026 | █████████████████████ 29,000-odd

(2021 starts in September and 2026 is still going, so the first and last bars are partial. The middle four are final.)

How five years of music got into one collection

The seam is this: 102,823 of those records were imported by malachite, a tool I built in November 2025 to import Last.fm and Spotify listening history into AT Protocol repositories. It turned one two days ago, which feels like the right week to finally write all this up. It worked through eleven client versions — v0.3.0 to v0.19.10 — resubmitting my Last.fm backup in batches until that pipeline died in July 2026.

The remaining 13,500-odd are live scrobbles from piper, the upstream Teal.fm scrobbler, submitted by an instance run by Matt. I forked piper to help with development occasionally rather than to self-host, and one of those contributions made it into a release: the Last.fm-compatible scrobbling API that shipped in piper v0.2.0 this month. Anything that still speaks Last.fm can now scrobble straight into my repo. There's a symmetry there I enjoy.

So the collection spans two listening lives. Underneath the imports is Spotify, my player from 2021 until early 2026, relayed through Last.fm. Underneath piper is Apple Music, my player since 1 August 2026 (first scrobble of the new era: Seb Lowe, "Paradise of Mine"). Two wrinkles stand out.

The first: Apple Music and I had actually met before. For six weeks in August and September 2024, on a free trial I entirely forgot about, I logged 852 plays there — and the very first was AURORA's "Running with the Wolves" at 05:29 on 28 August 2024. That song is now my third most-played track of all time. Some things are constants.

The second: when Last.fm broke on 19 July 2026 and piper wasn't yet running, there's a twelve-day hole with no plays logged at all — my only gap in five years. There's also exactly one record submitted by a different client entirely, plyr.fm/1.0: a 15 October 2026 test play of a track called "mathmusic hardtekk test", credited to one Ewan Croft. I make no apologies for that one.

Six eras of taste

2021: The eclectic start. Dagames, Arlo Parks, Nathan Evans sea shanties, YOASOBI, Lil Nas X, AJ Tracey. A snapshot of a music library with no unified theory behind it yet. My very first logged play was Chicane — which turns out to be foreshadowing, because electronic music never really left.

2022: The instrumental era. Epic Mountain, the band behind the Kurzgesagt soundtracks, plus Blockhead, OGRE Sound and VGR. I was listening to a lot of YouTube, and my scrobbles were basically a browser history in disguise.

2023: The animation music peak. Epic Mountain hit 1,485 plays in a single year, with CG5, Steven Universe, AURORA and OR3O close behind. If you want to know what the algorithm was serving me that year, this is it.

2024: The pivot. Ethan Bortnick and Night Club took over the top spots, Poor Man's Poison appeared, and quietly, with just 356 plays, Powerwolf entered the building. Nobody paid attention at the time. Least of all me.

2025: The metal year. Powerwolf went from 356 plays to 6,150, and my yearly total exploded to 42,722 — still my all-time record. Cjbeards, Moon Walker, Thomas Haines and The Stupendium filled out the ranks. This was also the year I built malachite, so the back half of 2025 is when my whole rescued history finally landed on my own PDS.

2026: The widening. Powerwolf still leads, but the picture has broadened: Moon Walker, hardtek (LABORATOR, Hardtekk, ZMAJOR & Lurk), and a genuine wave of sleep and ambient artists — The Deep Sleep Scientists, yoneiu, Amethyst Skies — with tracks literally named "calmative (sleep)". The music I use has started to look as varied as the music I love.

What the rankings say

Top artists, all time:

All rounded to the nearest hundred, because these are all still climbing — exact counts wouldn't survive the week.

Top tracks, all time (nearest ten, same deal):

And the most-played releases: The Sacrament of Sin (about 1,300 plays), Kurzgesagt Vol. 5 (about 1,100), Blessed & Possessed and the Wolf King Netflix soundtrack (both just over a thousand), and Wake Up The Wicked — just shy of a thousand, plus another eight hundred and change on the deluxe edition, because apparently I don't do things by halves.

An entirely reasonable number of wolves

Here's the part where I come clean: I ran a keyword sweep of every track, artist and album title in the collection, and more than 11,500 plays — almost exactly one in ten of everything I've ever logged — is wolf-themed. Roughly two-thirds of those are Powerwolf alone, a band sitting at about seven and a half thousand plays since November 2024, a number that refuses to hold still. But the rest isn't the band: it's "Running With The Wolves", it's "Wolf Blood", it's the Wolf King soundtrack sitting in my five most-played releases.

This isn't a coincidence the data uncovered — it's a coincidence the data confirmed. I'm a werewolf person; that will not be news to anyone who has so much as glanced at this blog. Lycanthropic horror has been my all-time favourite genre for going on fifteen years, and I wrote a whole essay about it earlier this month (Necranthropy: Zombies Are Werewolves Facing the Other Way, if several thousand words on monsters sounds like your kind of afternoon). I play a Wolfkind character in D&D and keep his in-character journal on my website. My site bio openly admits to an "entirely reasonable enthusiasm for werewolves". Even the little digital person I run is named faol — Gàidhlig for wolf. If anything, the surprise is that the wolves only account for ten percent.

Around the clock

The hour-by-hour histogram is the chart that surprised me most. There is no real quiet hour.

Plays by hour of day (UTC), rounded to the nearest hundred; bar length proportional to plays:

00 | █████████████████ 5,700
01 | █████████████████ 5,500
02 | ████████████████ 5,100
03 | ████████████ 4,000
04 | █████████ 3,000
05 | ████████ 2,600
06 | ████████ 2,600
07 | █████████ 2,900
08 | ████████████ 4,000
09 | ████████████ 3,800
10 | ████████████ 3,900
11 | █████████████ 4,100
12 | ███████████████ 4,800
13 | ██████████████████ 5,800
14 | ██████████████████ 6,000
15 | ████████████████████ 6,500
16 | ████████████████████ 6,400
17 | ██████████████████ 5,900
18 | ████████████████ 5,300
19 | █████████████████ 5,700
20 | ██████████████████ 5,800
21 | █████████████████ 5,700
22 | ██████████████████ 5,700
23 | █████████████████ 5,500

Listening peaks mid-afternoon around 15:00, bottoms out around 06:00, and never actually stops: midnight to 2am still accounts for over 16,000 plays across five years.

Oidhche mhath (good night)

Since the Apple Music switch in August 2026, my library has grown a quiet, droning wing, and it explains my two most extreme records. My busiest month so far is August 2026 with 11,902 plays, and my busiest day so far is 8 August 2026 with 719 plays in 24 hours — a day whose top artist was The Deep Sleep Scientists, with 64 plays of their own. Nobody consciously listens to 719 tracks in a day. The stereo does it, while I sleep. Wind-down music in the most literal sense.

Wednesdays are marginally my heaviest weekday, for what it's worth, and the spread between my busiest and quietest days is under a fifth. I don't really have listening days and non-listening days. I just have days.

Methodology

Every play lives in fm.teal.feed.play on my PDS at eurosky.social, attributed to did:plc:ofrbh253gwicbkc5nktqepol (@ewancroft.uk), published through Teal.fm. Imported records came from malachite, my Last.fm and Spotify history importer; live scrobbles come from piper, whose instance I use courtesy of Matt. I fork it to help with development occasionally, and my Last.fm-compatible scrobbling API shipped in its v0.2.0 release. For this post I paged through the entire collection via com.atproto.repo.listRecords — 1,165 paginated calls — and aggregated the results locally. Durations are present on about 23,000 records, all Apple Music era, which is why total listening time is an estimate; everything else is either exact and closed, or rounded and still climbing.

Final thoughts

Five years of data says I'm less a person with one genre and more a person with seasons: sea shanties, then YouTube instrumentals, then animation soundtracks, then a year where power metal devoured everything, and now a blend of metal, hardtek, dark folk and ambient music for sleeping — all of it recorded first by Last.fm and then, once I'd built the tools for it, by me, on infrastructure I own.

And through all of it, one in ten songs is about wolves — which isn't a quirk of the data. It's the best-documented enthusiasm I have.

If you want to see the raw history, it's all public. Point anything that speaks AT Protocol at my DID and the fm.teal.feed.play collection, and you can run the numbers yourself. Your totals will be higher than mine — the count moves on between my fetching it and your reading it, which is rather the point of a live log. It never sits still.

Comments

Checking AT Protocol sign-in…

No comments yet

Be the first to share your thoughts on this post.