Probabilistic modeling of NFL field goals
Research project using logistic regression, multiple linear regression, neural networks, and random forests to form probability estimates on NFL field goals. Using these estimates, an unbiased measurement of individual kicker ability can be formed by measuring points added above the expected value per attempt over individual seasons or entire careers. Presented at Boston College Big Data Research Day as "Machine Learning and the NFL Field Goal."
Interactive demo can be found at https://jamesledoux.shinyapps.io/fg-models/
Full writeup in the Field Goal Writeup.pdf file.