Predicting the Fantasy Points in a subsequent game for a NFL Running Back (RB) and Wide Receiver (WR) based off of the statistics from the previous game.
Presentation Link: YouTube
Features (Independent Variables):
- RB: Carries/Game,Yards/Carry, Rushing Yards, Rushing Touchdowns, Receptions, Receiving Yards, Receiving Touchdowns, Fumbles
- WR: Receptions, Receiving Yards, Receiving Touchdowns, Carries/Game,Yards/Carry, Rushing Yards, Rushing Touchdowns, Fumbles
Dependent Variable (Target): Fantasy Football Points for next game
- Analyze data from 2019 season from Pro-Football-Reference (https://www.pro-football-reference.com/)
- Running Backs and Wide Receivers around 500 Rushing/Receiving Yards and above
Apply all regression models as aforementioned and compare Train, Validation, and Test R^2 values to determine best for model. Utilize other regression models such as LASSO and Ridge to Feature Engineer. Create running average of statistics for entire 2019 season (ie.: If Game 3 and predicting Fantasy Points for Game 4, then Game 3 statistics will be an average of games 1, 2, and 3.)
Present model with the best RSME score to determine how many points model comes within actual score. Reduce RSME as much as possible to create most accurate model.
Aggregate Features on 4-week running averages (ie.: Game 5 features predicting Fantasy Points for Game 6, average statistics from Games 2,3,4, and 5)
- Fantasy Football Enthusiasts
- basics of the web (requests, HTML, CSS, JavaScript)
BeautifulSoup
(web scraping)numpy
,pandas
,Jupyter Notebooks
statsmodels
,scikit-learn
Seaborn
Matplotlib
Yellowbrick