The raw data behind the story "The Complete History Of The NBA" https://projects.fivethirtyeight.com/complete-history-of-the-nba/ and our "2017-18 NBA Predictions" https://projects.fivethirtyeight.com/2018-nba-predictions/

nba_carmelo

Format

A data frame with 69,465 rows representing games and 20 variables:

date

Date

season

Season year, 1947-2018

neutral

TRUE if the game was played on neutral territory, FALSE if not

playoff

TRUE if the game was a playoff game, FALSE if not

team1

The name of one participating team

team2

The name of the other participating team

elo1_pre

Team 1's Elo rating before the game

elo2_pre

Team 2's Elo rating before the game

elo_prob1

Team 1's probability of winning based on Elo rating

elo_prob2

Team 2's probability of winning based on Elo rating

elo1_post

Team 1's Elo rating after the game

elo2_post

Team 2's Elo rating after the game

carmelo1_pre

Team 1's CARMELO rating before the game

carmelo2_pre

Team 2's CARMELO rating before the game

carmelo1_post

Team 1's CARMELO rating after the game

carmelo2_post

Team 2's CARMELO rating after the game

carmelo_prob1

Team 1's probability winning based on CARMELO rating

carmelo_prob2

Team 2's probability of winning based on CARMELO rating

score1

Points scored by Team 1

score2

Points scored by Team 2

Source

See https://projects.fivethirtyeight.com/nba-model/nba_elo.csv