|
Abstract:
|
American football teams in the National Football League (NFL) spend vast resources every year scouting players at the collegiate level in preparation for the NFL draft. A successful draft class for a team often leads to success on the field for years to come, so the expenditure of resources is justified. While each team creates its own scouting report, experts employed directly by the NFL write their own reports which include a description of the player's strengths and weaknesses along with a player grade, a scale from 0 to 100 that allows direct comparison between players. We use that information, scraped from the World Wide Web, to determine if certain characteristics of players at different positions are more indicative of future success. We used natural language processing and machine learning methods to predict the Approximate Value metric (a measure of overall player value created by pro-football-reference.com) of players shortly after joining the NFL using the words contained in the NFL draft experts' reports.
|
