Predicting the 2024 AP NFL MVP Using Random Forest Regression
With the NFL regular season coming to an end, football fans and sports analysts alike have started to talk about the AP NFL MVP, and, specifically, how to predict it.
Here is my attempt to predict the AP NFL MVP for 2024.
Although I kept the NFL-MVP model almost exactly the same from last year (see results here), I made a change with how defensive players are considered in the model, namely whether to consider them at all.
This is a larger issue within the NFL, but defensive players are widely overlooked for the MVP award.
In the entirety of NFL history, only two defensive players have won the award: Alan Page in 1971 and Lawrence Taylor in 1986.
Although I disagree with the fact that defensive players are consistently overlooked for major awards in the NFL, I understand that the trend exists, and the model might produce more accurate results when only looking at offensive players.
While the final results included in the article only assessed offensive players, the version that considered offensive and defensive players is included in the Github repository.
Defensive players play a major factor in a team’s success and make the game so much more interesting and exciting to watch, and I hope they start to get the recognition they deserve.
I’m not sure I’ve ever heard a crowd roar louder than after someone gets a pick six.
I collected data using a web-scraping process (HTML files can be found in mvp, player, defense, and team folders) on mvp data, offensive and defensive statistics, and team data and cleaned and merged the data into a single csv file (stats.csv).
A huge shoutout to pro-football-reference.com for organizing an outrageous amount of NFL data in a systematic website.
Using the combined csv file, I used data from 1996-2023 to train a random forest regressor model which creates a series of decision trees and averages predictions from those trees to predict the AP NFL MVP for 2024.
Predicted results for 2024 (updated January 1, 2025, 5:30 pm EST):