National Football League Game Outcomes

Elo ratings
logistic regression
Bradley-Terry
State-space
NFL game outcomes during the Patrick Mahomes’ era (2018-present)
Author

Ron Yurko

Published

July 8, 2024

Motivation

Since Patrick Mahomes became the starting quarterback for the Kansas City Chiefs in 2018, the Chiefs have been arguably the best team in the NFL with three Super Bowl championships. This dataset is comprised of NFL game outcomes during the Mahomes’ era (2018-present) and can be used for building measures of NFL team strengths, such as Elo ratings.

Data

This dataset contains NFL game outcomes for every regular season and playoff game from 2018 to the present. The data were collected using the nflreadr package in R.

Each row in the dataset corresponds to a single game, with a total of 1,657 games and 10 columns containing relevant information about each game:

nfl_mahomes_era_games.csv
Variable Description
season Regular season year the game was played in
game_id Unique game ID based on the year, week, and teams
game_type String indicating the type of game, which is either regular season (REG) or in the playoffs (wildcard WC, divisional DIV, conference championship CON, and Super Bowl SB)
week Integer denoting the week of the season (note the playoffs week just continues from the end of the regular season)
home_team NFL team three letter abbreviation for home team
away_team NFL team three letter abbreviation for away team
home_score Number of points scored by the home team
away_score Number of points scored by the away team
game_outcome Indicator variable denoting if either home team won (1), away team won (0), or if the game ended in a tie (0.5)
score_diff Difference between home_score and away_score

Note: There are instances of rows where the Raiders are observed as either OAK or LV. An initial pre-processing step would be to convert all instances of OAK to LV.

Questions

  1. Use this dataset to build measures of team strength using an approach such as Elo ratings (SCORE Network module in progress, and will be available later).

  2. What is effect of home-field advantage in the NFL and how has it changed over time during the Mahomes’ era? You could use some type of regression technique (logistic for predicting win/loss, or linear for modeling the score differential) to infer the advantage (if any) a team has when playing at home.

  3. Join this dataset another dataset containing team statistics and build a logistic regression model to predict which team will win the game. What variables are most predictive?

References

Ho T, Carl S (2022). nflreadr: Download ‘nflverse’ Data. R package version 1.3.1, https://CRAN.R-project.org/package=nflreadr.