Introduction

Basketball games in the NBA are nice for a logistic regression approach because they only have two possible outcomes: you win or you lose. We dont have to deal with draws, even in the regular season.

We can use a individual team effects in a logistic regression approach to estimate the likelihood that a specific team gets a win against another given team. Clearly stated, we might ask based on the last 3 seasons of games, what’s the probability that the Toronto Raptors beat in the Golden State Warriors for a game played in the Bay Area?

Data

The nbastatR package for R is a conventient API wrapper to pull rich player and game statistics from the NBA Stats API and Basketball-Reference (as well as other sources for non-NBA basketball leagues). You can install the package through github with devtools.

We’ll load in the team and player reference tables, as well as game data for the past 3 seasons. We’ll then transform the data structure somewhat in preparation for analysis. To starts, let’s create dummy variables of the winners and losers for each game.

## Acquiring NBA basic player game logs for the 2018-19 Regular Season
## Warning: All elements of `...` must be named.
## Did you want `dataTables = c(typeSeason, dateGame, idGame, numberGameTeamSeason, nameTeam, 
##     idTeam, isB2B, isB2BFirst, isB2BSecond, locationGame, slugMatchup, 
##     slugTeam, countDaysRestTeam, countDaysNextGameTeam, slugOpponent, 
##     slugTeamWinner, slugTeamLoser, outcomeGame, namePlayer, numberGamePlayerSeason, 
##     countDaysRestPlayer, countDaysNextGamePlayer, idPlayer, isWin, 
##     fgm, fga, pctFG, fg3m, fg3a, pctFG3, pctFT, hasVideo, fg2m, 
##     fg2a, pctFG2, minutes, ftm, fta, oreb, dreb, treb, ast, stl, 
##     blk, tov, pf, pts, plusminus, fpts, urlTeamSeasonLogo, urlPlayerStats, 
##     urlPlayerThumbnail, urlPlayerHeadshot, urlPlayerActionPhoto, 
##     urlPlayerPhoto)`?
## Warning: `cols` is now required when using unnest().
## Please use `cols = c(dataTables)`
## Acquiring NBA basic team game logs for the 2018-19 Regular Season
## Warning: All elements of `...` must be named.
## Did you want `dataTables = c(typeSeason, dateGame, idGame, numberGameTeamSeason, nameTeam, 
##     idTeam, isB2B, isB2BFirst, isB2BSecond, locationGame, slugMatchup, 
##     slugTeam, countDaysRestTeam, countDaysNextGameTeam, slugOpponent, 
##     slugTeamWinner, slugTeamLoser, outcomeGame, isWin, fgmTeam, 
##     fgaTeam, pctFGTeam, fg3mTeam, fg3aTeam, pctFG3Team, pctFTTeam, 
##     hasVideo, fg2mTeam, fg2aTeam, pctFG2Team, minutesTeam, ftmTeam, 
##     ftaTeam, orebTeam, drebTeam, trebTeam, astTeam, stlTeam, 
##     blkTeam, tovTeam, pfTeam, ptsTeam, plusminusTeam, urlTeamSeasonLogo)`?
## Acquiring NBA basic team game logs for the 2019-20 Regular Season
## Warning: All elements of `...` must be named.
## Did you want `dataTables = c(typeSeason, dateGame, idGame, numberGameTeamSeason, nameTeam, 
##     idTeam, isB2B, isB2BFirst, isB2BSecond, locationGame, slugMatchup, 
##     slugTeam, countDaysRestTeam, countDaysNextGameTeam, slugOpponent, 
##     slugTeamWinner, slugTeamLoser, outcomeGame, isWin, fgmTeam, 
##     fgaTeam, pctFGTeam, fg3mTeam, fg3aTeam, pctFG3Team, pctFTTeam, 
##     hasVideo, fg2mTeam, fg2aTeam, pctFG2Team, minutesTeam, ftmTeam, 
##     ftaTeam, orebTeam, drebTeam, trebTeam, astTeam, stlTeam, 
##     blkTeam, tovTeam, pfTeam, ptsTeam, plusminusTeam, urlTeamSeasonLogo)`?
## Acquiring NBA basic team game logs for the 2020-21 Regular Season
## Warning: All elements of `...` must be named.
## Did you want `dataTables = c(typeSeason, dateGame, idGame, numberGameTeamSeason, nameTeam, 
##     idTeam, isB2B, isB2BFirst, isB2BSecond, locationGame, slugMatchup, 
##     slugTeam, countDaysRestTeam, countDaysNextGameTeam, slugOpponent, 
##     slugTeamWinner, slugTeamLoser, outcomeGame, isWin, fgmTeam, 
##     fgaTeam, pctFGTeam, fg3mTeam, fg3aTeam, pctFG3Team, pctFTTeam, 
##     hasVideo, fg2mTeam, fg2aTeam, pctFG2Team, minutesTeam, ftmTeam, 
##     ftaTeam, orebTeam, drebTeam, trebTeam, astTeam, stlTeam, 
##     blkTeam, tovTeam, pfTeam, ptsTeam, plusminusTeam, urlTeamSeasonLogo)`?

## Warning: `cols` is now required when using unnest().
## Please use `cols = c(dataTables)`

The Model

Hierarchical Bayesian Logistic Regression