Get NFL statistics of all games and players since 2012.
Data collected through this script includes information about potential response variables that could be interesting to the user, including:
- which team won the game
- whether or not the favored NFL team beat the spread
- whether the teams involved in each game scored more points than the over/under
Predictor variables that may be useful were also collected. I won't mention examples here because there are numerous team- and player-level statistics available. For the full data sets, check out the corresponding page within the project. Below are summaries of each file in this folder:
game_summary_by_team: One row per team, per game played; each game has two rows (one for each team involved in the game). Columns withteam_prepended are statistics regarding the particular team, while columns withopp_prepended are in regards to the opponentgame_summary_by_player: One row per game, per player that participated in the game. This captures individual game statistics for every player. If a player has played 16 games in his career, he would have 16 rows in this CSVplayer_metadata: For every player included in thegame_summary_by_playerCSV, "metadata" about that player - Height, weight, position, and metrics from their rookie combine. One row per player.
This particular project was conducive to using python classes because of how hierarchical the relevant concepts are. Seasons have Weeks, Weeks have Games, Games have Teams, Teams have Players, and Players have Positions.
Each one of these objects is a separate class (taking methods/attributes of parent classes, as appropriate); all classes can be found here
-
Folder structure based on article found here: https://www.kdnuggets.com/2018/07/cookiecutter-data-science-organize-data-project.html
-
Scripts to collect data:
src/data(output goes todata/external) -
Additional steps: I'm hoping to work on models which can somewhat-reliably predict the response variable(s) specified above. More to come!