Methodology

Scope and sources

Encore follows 7 bands (Arctic Monkeys, Oasis, Linkin Park, Twenty One Pilots, Muse, Metallica, Avenged Sevenfold), fixed for this version. Setlists (which songs were played at which shows) come from setlist.fm; the discography (albums, singles, tracks and release dates) comes from MusicBrainz.

Everything here is computed from 8,591 shows and published only as aggregates: by band, tour, year and album.

What the numbers mean

• Match to MusicBrainz. A performance (one time a song was played at a show) is matched when its title resolves to a song in the MusicBrainz discography. The share of matched performances, weighted by how often each song was played, is the main data-quality measure (table below). The share matched to a studio album is shown next to it for information only.
• What is left out. Covers, tape entries and intros are excluded from the measures. Medleys (titles joined by ' / ') are split into separate songs.
• Repertoire age. For each performance, the year of the show minus the song's first official release year (the earlier of its studio album's year and its recording's first release), never below zero. A year's figure is the mean over its performances. On the tour chart each bar is the median of one tour in one year, because a median cannot be rebuilt across several years.
• Rotation. For consecutive shows of the same tour, the overlap between the two setlists (Jaccard index) is computed. Rotation is 1 minus the mean overlap: 0 means the same setlist every night, 1 means no song in common. Tours with fewer than 5 shows are not scored, and a pair belongs to the year of its second show. Years with fewer than 5 pairs are drawn as hollow markers and left out of the line.
• Album of a song. The earliest studio album that contains it. Songs matched to MusicBrainz but not to a studio album (singles, B-sides, other recordings) form their own group, non-album, instead of being dropped.

Song survival and the abandonment rule

Survival is measured in shows, not in calendar time, so that a hiatus is not read as abandoning songs. A song's duration runs from its first show to its last, inclusive: a song played at one show only lasts 1.

A song is abandoned when it left the setlist and did not come back: its last appearance is at least N shows before the end of the history, with N = 50 by default. Every other song is censored: it is still being played, or it disappeared for a while and returned. Only discography songs (those found in MusicBrainz) that were played at least 3 times in total, matched to a studio album or to a recording with a known release year, enter the analysis.

Only the final gap counts. An earlier version treated the first gap as the event and marked classics that were rested for a tour and brought back as abandoned; the final-gap rule fixes that.

Comparisons between bands are made at a common horizon, 500 shows after a song's live debut, which every band's curve reaches, and not at the end of each curve: how far a curve extends depends on how long the band was followed, so a band with a longer career would end lower for that reason alone. The end of each curve is still given per band, with the number of shows it covers.

How much does N matter?

Median survival, in shows, of each band's eligible songs for three window sizes. Where the median barely moves, the choice of N is not driving the result; where it moves a lot, read the figure with that in mind.

Not reached: the curve never drops below 50%.

Data policy

Raw setlist.fm data exists only while the pipeline runs and is deleted at the end of every run. Only aggregated results are kept and published, so this site never shows an individual setlist, show date, venue or setlist.fm identifier.

Song titles, where they appear, are canonical MusicBrainz titles (CC0). Setlist data is credited to setlist.fm and used for non-commercial purposes only; discography data comes from MusicBrainz.

Known limitations

• A window in shows is not a window in time. Fifty shows pass in months for a band that tours heavily and in more than a year for one that does not. From 2015 on, an average active year has about 64 show pairs for Twenty One Pilots (its busiest year has 127), 52 for Arctic Monkeys, 40 for Muse, 40 for Oasis, and 28, 31 and 30 for Metallica, Avenged Sevenfold and Linkin Park (pairs slightly undercount shows). Compare bands with this in mind.
• Muse, 1994 and 1995. The match to MusicBrainz is low in Muse's two earliest years: 0 of 9 performances matched in 1994 and 12 of 50 in 1995 (24.0%). Together they are 0.25% of Muse's performances, so the overall match (99.0%) barely moves, but Muse figures for those years are thin. Likely cause: early demos and EPs that are not in MusicBrainz.
• Jams, solos and generic entries such as '(jam)' cannot be told apart reliably in the data, so where they match a discography song they are counted like any other. This can enlarge setlists, lower the overlap between shows and add spurious songs to survival.
• Recordings without a release year cannot be placed on the discography timeline and are left out of survival. Recording-only songs that do have a year are included under non-album.
• Release dates come from MusicBrainz and can be wrong. Suspicious cases are reviewed and corrected through a manual override list in the repository.

Match to MusicBrainz by band

Share of performances matched to MusicBrainz, weighted by plays (the main measure), and the share matched to a studio album (secondary, informational).