Here is the Shiny App. Still working on a better way to have it display within the website without truncating the display. Below is the trunctuated program, you can see the full program here.
Background
Last year I started this tool to explore how player projections line up with expert predictions. For example, Sammy Watkins is the 17th ranked WR per Fantasy Pros Expert Consesnus Rankings, but by consensus statsitical projections he’s the 10th best.
Ideally, this app should show you what players are over/undervalued for their draft slot/ranking.
It’s still very much in a beta phase, particularity in the regards to the ugly, ugly, UI that’s present right now. Content wise, it’s just about done, so I wanted to post it, and at least have the data available. . I’m planning on adding a table of all the data below a given graph, so you can easily export that to Excel. If there’s any other content you’d like added, please leave a comment.
Methodology
Some notes on the methodology used:
- I used a K-Means clustering algorithm to sort the players into distinct tiers.
- Based on my (limited) understanding K-Means isn’t generally used for 1-D clustering as I’ve done here, but I followed the methodlogy used in this article and felt K-Means was a easy way for me to implement tiers
- All the code was written in R. I used Plotly to help me generate the chart and RStudio’s Shiny to host and create the interactive sliders.
- All projections are from FantasyPros consensus rankings
- The “baseline” checkbox is based on this value based drafting guide, where the worst starter is used to determine the value of all other positions.