Shakedown is not just a place to see the same song titles crammed into a popular logo or watch zombies flock to the sound of a hissing tank. Shakedown is also a parking lot. And a parking lot isn’t just for cars. A parking lot can be a place to park ideas and questions that come up but arne’t a priority in the moment. Below I’ll just dump all the data questions and project Ideas I get along the way.

Blogs Posted

• Why Phish Data? - Why did I choose to use Phish data to teach me data science concepts?
• Anatomy of a Phish Stats Post - The structure I’ll use to post Phish stats
• Day of week bias analysis - Are some songs played more frequently on certain days of the week?
• What Do I Need to Know About the Data? - I’m using a database from Phish.in that is focused more on uploading tracks and less on analyzing Phish data so I need to see what quirks are in the data and how to compensate in different projects.
• Song frequency changes over the course of Phish’s career
• Visualizing Changes in Number of Songs per Show - A handful of visualizations showing how the number of songs in each set and in a full show have changed over time.
• Analysis Set Placement of Specific Songs - Which songs tend to be openers or closers? Which songs tend to be played in the same slot and which ones get spread throughout a show?
• Dinner and a Movie - 7/21/97 - Visualizations - A visualization of a Phish show.
• Dinner and a Movie - 12-29-18 - Visualizations - A visualization of a another Phish show.
• Predicting Phish Set Closers - An interactive dashboard where you can input any Phish show and it will tell you the liklihood of each track being the last song of the set.
• Phish Interactive Dashboards - Links with explanations to the three flask/Dash apps I made for Phish data.

Projects Almost Done

• How have Halloween albums impacted future setlists - Publish along with similar analysis for album releases

Project Ideas

• Song cluster analysis - Which songs are most likely to show up in the same set or show. (This is part of the “story of a Phish song” dashboard but I plan to add more data)
• Are there patterns on when a song will turn up during a multi-night run
• Mostly Mike, Prolific Page, Frequently Fish, and Totally Trey shows - Which shows have the greatest percentage of songs sung by each member.
• Redefining Phish eras by significant changes in song lengths, show lengths, song selection, etc
• Tagging specific types of showsf - Seguefests for songs with x number of segues, number of bustouts, number of tracks
• Song length and spread analysis - Do songs tend to get longer or shorter over time, can you predict how long the next song will be based on the length of the previous track? How much is the length of the longest song in a set correlated to the number of tracks in that set?
• Light Color Analysis - Comparing all official youtube releases, or potentially dinner and a movie releases, to look for patterns in light colors. Like a combination of these two projects:
• State or venue bias - Are they more likely to play certain songs in certain states or venues?
• Predicting the frequency of a song based on its debut - Is it possible to predict how frequently a new song will be played based on the set placement of the debut, or the length of the debut, or how far into a tour the song is debuted?
• Predicting the distribution of how many times each song will be played in the next 50 shows.

Requests

Apart from focusing on the projects that get me to practice skills connected to machine learning, I don’t really have a preferences for which of these projects I work on next. So, requests are welcome (jroefive@gmail.com), either from the list above or ideas you might have. Just don’t hold you sign up for too long and block the view of the people behind you.