Somebody Else Already Chose: Why the Compilation Beats the Album¶
The Pattern¶
Split every rated track in your library by whether it came off a compilation — a best-of, a retrospective, an anthology, a De Afrekening volume, a scene sampler — or off a studio album.
| mean | 4★ rate | n | |
|---|---|---|---|
| compilations | 3.37 | 44.6% | 654 |
| studio albums | 3.20 | 34.0% | 6,755 |
Ten and a half points of hit rate. For comparison, that gap is wider than the one between your best genre lane and your worst. It is one of the largest effects anywhere in the model, and it has been sitting in plain sight across 7,409 ratings.
The top of the compilation table is not marginal material:
| album | rated | mean |
|---|---|---|
| Meet the Eels: Essential Eels Vol. 1 | 22 | 4.14 |
| In Time: The Best of R.E.M. 1988–2003 | 8 | 4.00 |
| Neil Young — Decade | 13 | 4.00 |
| Urban Revolutions: Future Primitive Sound | 15 | 3.87 |
| The Very Best of Power Ballads | 6 | 3.83 |
| Party in Session: The Black Uhuru Collection | 7 | 3.71 |
| U2 — The Best of 1990–2000 | 10 | 3.70 |
| De Afrekening 15 | 6 | 3.67 |
| Wave of Mutilation: Best of Pixies | 15 | 3.47 |
| Retrospective: The Best of Suzanne Vega | 18 | 3.39 |
Meet the Eels at 4.14 is the highest-rated album of any kind in your collection with a meaningful sample. And The Very Best of Power Ballads: The Greatest Driving Anthems …Ever! at 3.83 outrates most of your post-metal — which is the sort of thing a taste model will tell you if you let it, and a self-image will not.
De Afrekening alone accounts for 176 rated tracks at 3.27 with 63 four-stars, spread across at least a dozen volumes.
Historical Context¶
Two separate mechanisms are producing this number and they are worth keeping apart.
The first is straightforward: a best-of is a filtered object. Somebody at a label, or the artist, already removed the weak tracks. Comparing Meet the Eels to Daisies of the Galaxy is not comparing two records, it is comparing a record to a record with its worst third deleted. That mechanism is real, it is boring, and it explains a good part of the gap.
The second is not boring, and it is specific to you. Your formation ran through compilations as a primary medium rather than as a convenience.
De Afrekening was Studio Brussel's alternative chart, running from the late 1980s, compiled annually onto CD. For a Flemish teenager in the 1990s that series was not a shortcut to an artist's catalogue — it was the catalogue. It was how you learned that Nick Cave and the Pixies and Belgian guitar bands belonged to the same conversation. The station was a publicly-funded broadcaster with editorial independence, so what got on those discs reflected the judgement of people paid to have judgement, not a label's promotional cycle. That is the same institution that produced Duyster.
Then Napster, which the self-portrait names directly as the origin of the ownership instinct — necessity first, habit second. The unit of acquisition in 1999 was the track, not the album, and the object you ended up with was a folder somebody had assembled. Urban Revolutions: The Future Primitive Sound Collective (3.87) already has its own essay in this corpus for exactly this reason.
So the compilation is not a format you tolerate. It is the format your listening was built in, at precisely the age when — as the self-portrait puts it — everything was felt more acutely. The album-as-artwork orthodoxy, which the same document elsewhere defends as objectively better music, is a value you hold; the compilation is how the music actually reached you. Those are different things and the ratings can tell them apart.
There is a third possibility worth naming rather than smuggling in: rating
bias. A best-of contains the songs you already know, and familiarity may
simply pull the rating up. That is not separable with the data available —
play_count is unusable, with only fifty rated tracks above eight plays and
most of those from the family playlist. It is a real confound and this
finding should be held with that caveat attached.
The Recordings¶
The strongest version of the pattern shows up where you own both forms of the same artist.
eels. Meet the Eels rates 4.14 across 22 tracks. The studio albums it draws from — Beautiful Freak, Electro-Shock Blues, Daisies of the Galaxy — sit lower individually, and the artist overall is 4.00 on interior-tagged material. Two songs, "Flyswatter" and "Last Stop: This Town", appear at five stars twice: once on the studio album, once on the compilation, rated independently. That duplication is a quiet consistency check on your own ratings, and it passed.
R.E.M. In Time rates 4.00 across 8 while the artist overall sits at
3.56 across 43 — a point already made by the rock-early-1990s essay,
which read it as a tell that this is singles-based listening rather than
deep-catalogue immersion. The wider data now says that is not an R.E.M.
fact, it is a library-wide one.
Neil Young — Decade at 4.00 is the interesting case, because Young is a confirmed pillar at 3.85 across 39 rated with a large catalogue. Even for an artist you demonstrably go deep on, the curated triple-album still beats the studio average.
Pixies are the sharpest illustration. The artist overall is 3.05 across 74 rated tracks — one of the flattest means of any major artist in your collection. Wave of Mutilation: Best of Pixies is 3.47. Same band, same songs, 0.42 apart, entirely because of what got left off.
What's Missing¶
The operational read matters more than any purchase here, and it is slightly uncomfortable: the acquisition pipeline is optimised to bring in albums, and the compilation result says the album is the weaker unit for you by ten points of hit rate. Meanwhile the playlist system — Similarity Pool, For Life lanes, the sweeps — is in effect a machine for generating new De Afrekening volumes. That is the part of the system best aligned with how you actually listen, and it is the part that gets described as infrastructure rather than as the point.
The gaps below are compilations proper: scene documents rather than best-ofs, in lanes that already rate well, all verified absent.
Acquisition Targets¶
- Various — Trojan Dub Box Set (1998) — the dub survey your roots lane lacks; you own the artists individually and none of the canonical overview compilations.
- Various — Nuggets: Original Artyfacts from the First Psychedelic Era (1972/1998) — the archetype of the curated scene document, and the 1960s garage material is a genuine hole given how thickly 1990s guitar music is covered.
- Various — Nigeria 70: The Definitive Story of 1970's Funky Lagos (2001) — Afrobeat and highlife as a curated entry point, adjacent to the Latin/World lane that Buena Vista Social Club already anchors at 3.54.
- Various — Ocean of Sound (1996) — David Toop's ambient survey, which is the compilation equivalent of the Duyster programme and would test that lane as an environment rather than as tracks.