Image credit: Jarrod Barker
“You Don’t Have to Count Them, Just Enjoy Them One by One”
– Quantifying Phish
Paul Jakus, “Bombers and Fluffers: Anomalous Rating Behavior and Show Rankings”
Using an (anonymized) Phish.Net ratings database composed of over 16,000 people providing over 340,000 show ratings, each rater/user is evaluated according to metrics and methods grounded in information theory and marketing science. Metrics include the number shows rated by a user, deviation in ratings between the user and other users, and entropy of a given user’s ratings. Entropy—a measure of information provided by a user’s ratings—is, by far, the most useful metric in identifying anomalous users. In contrast to the current .Net rating system, which applies an equal weight to all users, one can use the above metrics to vary weights across users. Show rankings under the current system are compared to show rankings generated by five alternative weighting systems, including one designed to mimic the system used by rateyourmusic.com.
Matt Sottile, “A Quantitative Examination of Phish History Through Audio Recordings”
The presence of tapers and the surrounding trading scene has existed for the bulk of Phish’s career resulting in a detailed performance history of the band via easily accessible audio recordings. This allows us to examine the content of the performances to look at how Phish and their repertoire has evolved. In this talk I will discuss the kinds of quantitative information that can be derived from this recorded history both manually and automatically. To illustrate some of the results I will take one of the band’s longest standing songs as an example of study: Reba. I will look at Reba from a statistical and signal processing perspective to see what can be teased out of over 350 recorded performances of the song from 1989 through the present. I will also discuss the challenges that arise when applying audio analysis techniques to fan recordings.
Jason Zietz, “Who Does Phish Play For? Themselves? The Fans? A Little from Column A, A Little from Column B?”
While Phish’s setlists were originally predetermined, the more recent approach reportedly finds the band creating a list of many possible songs before the show begins, only to have Trey Anastasio, Phish’s guitarist and lead singer, destroy the list right before the show begins. Given this semi-random approach to setlist design, is it possible to determine what motivates the band to choose the songs they play?
In this paper, I provide the details of a quantitative analysis of various data associated with Phish’s recorded music and live performances that addresses this question. This analysis consists of several parts. First, I examine what songs, on average, fans of the band tend to prefer by determining the most popular songs on popular streaming services. I then analyze various qualities of these songs, including lyrical sentiment and audio properties such as key tonality, duration, and complexity, to determine if there are any commonalities among these most-liked songs.
To determine what the band primarily likes, I look at the frequency of songs played at shows and generate from the most frequently played songs some general criteria that may predict what the band chooses to play based on the songs’ audio qualities. Additionally, I also specifically analyze the cover songs the band most frequently plays, since learning another musician’s work likely indicates a high level of appreciation for that music.
Finally, I examine fan perceptions of shows to determine to what extent fan satisfaction correlates with fan preference of particular songs. This is done by examining both show ratings and semantic analyses of show reviews. These analyses are presented for the entire career of the band as well as segmented into the commonly used numerical “eras” (i.e, 1.0, 2.0, et cetera) to determine if, over time, any differences among these eras can be ascertained. These analyses help provide evidence that show setlists indicate that, rather than solely playing for either the band or the fans, Phish instead typically plays for, to quote Grandpa Abe Simpson, “A little from column A, a little from column B”.