I've been interested in the NFL Draft for a long time and finally decided to make a data science project out of it. I've always wondered what the point of mock drafts were since I don't know what makes one mock draft better than another or one "expert" better than another so I decided to try to answer those questions with data!
I've collected data on hundreds of mock drafts from the last two years (mostly by hand unfortunately) and built a data product around a set of questions:
- What are the consensus top prospects at each position by average, median, and a weighted (by how close a pick is made to the day of the draft) average mock draft position metrics.
- How are draft eligible players' stocks changing as the draft process goes on? Is a player's stock really going up or is that just hearsay? I want to follow along the data journey of a player's stock using a variety of methods to fit a curve to estimate where a player might fall and how their stock is trending.
- How well do mock drafts do on the whole for predicting where a player might get taken in the draft? Which players were the biggest "reaches" and "steals"?
- Finally, who are the true experts? Do fans, the media, or experts make the best predictions? I'm not as interested in whether an individual makes the best predictions but rather which class of mock drafters does the best. Which group's predictions should I be following to get the best picture of the NFL draft.
Looking forward to getting more feedback and to get this work out there...plus I can't lie having this app available more than just 25 active hours a month would be pretty clutch!
By the way, the data for this project is publicly available in case people want to use it for their own machine learning or data science projects!
ShinyApps.io Link: https://benjrobinson2.shinyapps.io/2018_NFL_Mock_Draft_Performance_Tracker/
RStudio.Cloud Project Link:
- Ben Robinson.