Visualizing NBA Seasons: Introduction & Preparation

Seasons 1950-2017

Introduction & Preparation

Part 1: Players

Part 2: Points

Part 3: Shootings (FG and FT)

Part 4: Shootings (3-Points and Mixed)

Part 5: Skills

Part 6: Roles

---

Introduction

This in an exploratory analysis of the NBA Seasons 1950-2017. If you are familiar with NBA statistics terminology, don’t care about how the data being processed, or just want to go straight to the fun, please skip this boring part and proceed to the next part.

About Me and the Project

It was the NBA who first introduced me into the broad world of statistics in an entertaining manner back in my childhood. Since then I have always been captivated by statistics and data.

Despite my interest in statistics coupled with intermittently making amateurish statistical data analysis out of anything that piqued my interest for pleasure, I never really had the proper training nor consider take the subject as a career until recently. It was not long ago, without prior knowledge in the area, I started to teach myself coding in Data Science obsessively by dive into MOOC (Massive Open Online Course) courses religiously in my spare time.

Being absent from following NBA regularly for a long time elevated my excitement to explore, analyze and gain insight from the data. Thus, this dataset is perfect for my first attempt at doing a self-directed project, as it is convenience for me to dig deeper into learning Data Science while reaping the joy of producing it.

I expect to use this opportunity to gain more experience by challenging myself to implement my new skill while expanding it by exploring various methods of data wrangling and visualization, then communicating the findings efficiently in an aesthetic manner. In this way, I can expect to shape my skills through practical learning.

The objective is to present NBA seasons statistical facts through compelling visualization and tables. I do not make any predictive analysis here (will get there later).

This presentation is designed for anyone who interested in this subjects without the need of prior technical knowledge, hence the simplification and references, it also intended to draw expertise in the subject to give feedback for my improvement.

I hope this may be useful as a reference for understanding the NBA statistics, NBA seasons history, or as a code reference for R learners

The Dataset

The dataset contains aggregate individual statistics for 67 NBA seasons from 1950 to 2017. The data was scraped from Basketball-reference and is outdated. Nevertheless, the codes are designed to facilitate update once new data emerge. . For simplicity, I took only the basic stats (FG, PTS, etc.), and dismissed the advanced stats (VORP, PER, OBPM, USG%, etc.).

Since the dataset provides regular season statistics and basic player biodata only, I need to point out the dataset limitation:

• No NBA champion, playoffs or All-Star is present in the dataset.
• There’s no win-lose nor score information.
• No team’s stats given.
• Data about title or awards (e.g.: MVP contest winners, rookie of the year, All-Star player, etc.) is not provided in the dataset.
• The player’s stats are presented as a whole season, so it’s not possible to extract information about in-game records.
• No regional division information is given.

---

Glossary

• Year: Since the NBA season is split over two calendar years, the year given is the last year for that season. So, the year for the 1999-00 season would be 2000.
• Player: A player’s name.
• Pos: Position (Center, Power Forward, Small Forward, Shooting Guard, Point Guard)
• Age: Player’s age on February 1st of the given season.
• Tm: The team, in which player rostered in a given season. A player can have more than a team in a season due to the frequent trade. For further information about team abbreviations and history visit this link.
• G (Games): Number of games played in a season
• GS (Game Started): Number of games played in a season as a Starting Lineups (available since the 1982 season)

• MP (Minutes Played): Total minutes of play in a season (available since the 1951-52 season)

• Born: A year that a player born.
• YearStart: Minutes of played in a season.
• YearEnd: Minutes of played in a season.
• Height: Minutes of played in a season.
• Weight: Minutes of played in a season.

• BMI (Body Mass Index): Minutes of played in a season.

• FG (Field Goals): Field Goals (includes both 2-point field goals and 3-point field goals)
• FGA (Field Goal Attempts): Field Goal Attempts (includes both 2-point field goal attempts and 3-point field goal attempts)
• FG. (Field Goal Percentage): Field Goal Percentage
• X3P (3-Point Field Goals): 3-Point Field Goals (available since the 1979-80 season in the NBA)
• X3PA (3-Point Field Goal Attempts): 3-Point Field Goal Attempts (available since the 1979-80 season in the NBA)
• X3P. (3-Point Field Goal Percentage): 3-Point Field Goal Percentage (available since the 1979-80 season in the NBA)
• X2P (2-Point Field Goals): Field Goals minus 3-Point Field Goals
• X2PA (2-Point Field Goal Attempts): Field Goal Attempts minus 3-Point Field Goal Attempts
• X2P. (2-Point Field Goal Percentage): 2-Point Field Goal Percentage; the formula is 2P / 2PA
• eFG. (Effective Field Goal Percentage): This statistic adjusts for the fact that a 3-point field goal is worth one more point than a 2-point field goal. For example, suppose Player A goes 4 for 10 with 2 threes, while Player B goes 5 for 10 with 0 threes. Each player would have 10 points from field goals, and thus would have the same effective field goal percentage (50%). The formula is \(\frac{FG + 0.5 * 3P}{FGA}\)
• FT (Free Throws): Free Throws
• FTA (Free Throw Attempts): Free Throw Attempts

• FT. (Free Throw Percentage): Free Throw Percentage; the formula is FT / FTA.

• ORB (Offensive Rebounds): Offensive Rebounds (available since the 1973-74 season in the NBA)
• DRB (Defensive Rebounds): Defensive Rebounds (available since the 1973-74 season in the NBA)
• TRB (Total Rebounds): Total Rebounds (available since the 1950-51 season)
• AST (Assists): Assists
• STL (Steals): Steals (available since the 1973-74 season in the NBA)
• BLK (Blocks): Blocks (available since the 1973-74 season in the NBA)
• TOV (Turnovers): Turnovers (available since the 1977-78 season in the NBA)
• PF (Personal Fouls): Personal Fouls

• PTS (Points): Points

• TS. (True Shooting %): Measure of shooting efficiency that takes into account field goals, 3-point field goals, and free throws. Equation: \(\frac{PTS}{2TSA}\)
• X3PAr (3-Point Attemp Rate): Measure of what % of a player’s shots come from long-distance, another good gauge of how they’re utilized offensively.
• FTr (Free Throw Rate): The ratio of free throws attempted per field goal attempted.
• ORB. (Offensive Rebound Percentage): Estimate of the percentage of available offensive rebounds a player grabbed while he was on the floor.
• DRB. (Defensive Rebound Percentage): Estimate of the percentage of available defensive rebounds a player grabbed while he was on the floor.
• TRB. (Total Rebound Percentage): Estimate of the percentage of available rebounds a player grabbed while he was on the floor.
• AST. (Assist Percentage): Estimate of the percentage of teammate field goals a player assisted while he was on the floor.
• STL. (Steal Percentage): Estimate of the percentage of opponent possessions that end with a steal by the player while he was on the floor.
• BLK. (Block Percentage): Estimate of the percentage of opponent two-point field goal attempts blocked by the player while he was on the floor.

• TOV. (Turnover Percentage): Estimates the number of turnovers a player commits per 100 possessions.

More about Positions

Historically, only three positions were recognized (two guards, two forwards, and one center) based on where they played on the court: Guards generally played outside and away from the hoop and forwards played outside and near the baseline, with the center usually positioned in the key. During the 1980s, as team strategy evolved after the three-point field goal and the three-point arc were added to the basketball court, more specialized roles developed, resulting in the five position designations used today. (source)

In here, for simplicity and esthetical purpose, I arbitrarily framed up all the players into the standard five position used today and then color-coded them. This is essensial because I tend to make good use of this for comparison frequently.

For more information about each position, click on the position icon below.

• : Center
• : Power Forward
• : Small Forward
• : Shooting Guard
• : Point Guard

---

Data Preparation Stage

Loading necessary packages.

library(dplyr)
library(tidyr)
library(measurements)
library(sqldf)
library(kableExtra)

Create additional functions

# Mode average
getmode <- function(v) {
   uniqv <- unique(v)
   uniqv[which.max(tabulate(match(v, uniqv)))]
}

convertHeight <- function(x) {
    x <- as.character(x)
    split <- strsplit(x, "-")
    feet <- as.numeric(split[[1]][1])
    inch <- as.numeric(split[[1]][2])
    round(conv_unit(feet, "ft", "cm") + conv_unit(inch, "inch", "cm"),0)
}

Load the dataset

Download and extracting the files.

Dataset from: https://www.kaggle.com/drgilermo/nba-players-stats/version/2

fileURL <- "https://www.kaggle.com/drgilermo/nba-players-stats/downloads/nba-players-stats.zip/2"
filename <- "NBASeason1950-2017.zip"

# Checking if archieve already exists.
if (!file.exists(filename)){
  download.file(fileURL, filename, method="curl")
}  

# Checking if folder exists
if (!file.exists("NBA Season Dataset")) { 
  unzip(filename)
}

Download date: 28 July 2018

Load the data.

NBA <- read.csv("NBA Season Dataset/Seasons_Stats.csv")[, c(2:9, 11:20, 32:53)]
PlayerData <- read.csv("NBA Season Dataset/player_data.csv")[, c(1:3, 5:6)]

Tidying data

Clean up data.

# Remove NA rows
NBA <- NBA %>% filter(!is.na(Year), !is.na(Player))
# Remove Team = TOT (which indicates total, when player played in more than 1 team in a season)
NBA <- NBA[NBA$Tm != "TOT",]
# Remove of "*" which indicates a player is a member of NBA Hall of Fame
NBA$Player <- gsub("\\*$", "", NBA$Player)
# Fix player data
PlayerData[2143, 4] = as.factor("6-2")
PlayerData[2143, 5] = 190
NBA[21304, 3] = "SG"

Renaming some columns of the dataset

colnames(PlayerData) <- c("Player", "YearStart", "YearEnd", "Height-feet", "Weight-lbs")

Merging data

NBA <- sqldf("SELECT * FROM NBA JOIN PlayerData ON NBA.Player = PlayerData.Player 
      WHERE NBA.Year >= PlayerData.YearStart AND NBA.Year <= PlayerData.YearEnd")

Fixing Position

In here, for simplicity and esthetical purpose, I arbitrarily framed up all the players into the standard five position used today and then color-coded them:

• “C-F” in original dataset becomes “C”
• “F-C” in original dataset becomes “PF”
• “F” in original dataset becomes “PF”
• “F-G” in original dataset becomes “SF”
• “G” in original dataset becomes “SG”
• “G-F” in original dataset becomes “SF”

NBA$Pos[NBA$Pos == "C-F"] <- "C"
NBA$Pos[NBA$Pos == "F-C"] <- "PF"
NBA$Pos[NBA$Pos == "F"] <- "PF"
NBA$Pos[NBA$Pos == "F-G"] <- "SF"
NBA$Pos[NBA$Pos == "G"] <- "SG"
NBA$Pos[NBA$Pos == "G-F"] <- "SF"
NBA$Pos <- factor(NBA$Pos, levels = c("C", "PF", "SF", "SG", "PG"))
PosColorCode <- c("C"="#FF0000", "PF"="#FFA500", "SF"="#DDDD00" ,"SG"="#0000FF", "PG"="#32CD32")

Create new variables:

Player info dataset

• Height: Convert unit from feet to cm
• Weight: Convert unit from lbs to kg
• BMI: Calculate Body Mass Index
• Born: Year of birth
• ORpG: (Offensive Rebounds per Game): Average offensive rebounds a player made in a game. (ORB/G)
• DRpG: (Defensive Rebounds per Game): Average defensive rebounds a player made in a game. (DRB/G)
• RpG (Rebounds per Game): Average rebounds a player made in a game. (TRB/G)
• ApG (Assists per Game): Average assists a player made in a game. (AST/G)
• SPG (Steals per Game): Average rebounds a player made in a game. (STL/G)
• BPG (Blocks Per Game): Average rebounds a player made in a game. (BLK/G)
• TPG (Turnovers Per Game): Average turnovers a player made in a game. (TOV/G)
• PpG (Points per Game): Average points a player made in a game. (PTS/G)
• Position: Same like Pos, but prettified

NBA <- NBA %>%
    rowwise() %>%
    mutate(Height = convertHeight(`Height-feet`),
           Weight = round(conv_unit(`Weight-lbs`, "lbs", "kg"),0),
           BMI = round(Weight / (Height / 100)^2, 2),
           Born = Year - Age,
           ORpG = ORB / G,
           DRpG = DRB / G,
           RpG = TRB / G,
           ApG = AST / G,
           SpG = STL / G,
           BpG = BLK / G,
           TpG = TOV / G,
           PpG = PTS / G,
           Position = cell_spec(Pos,
                            color = "white",
                            align = "c",
                            background = factor(Pos, c("C", "PF", "SF", "SG", "PG"),
                                                PosColorCode)))

Arrange the table.

NBA <- NBA %>%
    select(Year:MP, YearStart:BMI, "Born", FG:PTS, TS.:TOV., ORpG:PpG, "Position", -c("Height-feet", "Weight-lbs"))

Create duplicate table with normalized stats.

NBA_Scaled <- NBA %>% mutate_at(vars(c(G:MP, FG:PpG)), scale)

NBA_pM <- NBA %>%
    mutate_at(vars(c(FG:PTS)), function(x){ x/NBA$MP}) %>%
    rename_at(vars(c(FG:PTS)), ~paste0(.,"_pM"))

NBA_Scaled <- cbind(NBA_Scaled, NBA_pM[,c(15:36)])

Displaying raw tidy data table

NBA

• Number of rows: 22345
• Number of columns: 56
• Number of players: 3889
• Number of teams: 68
• File Size: 7275.5 Kb

Write tidy dataset for later use…

write.csv(NBA, file = "NBA_TidySet.csv")
write.csv(NBA_Scaled, file = "NBA_Scaled_TidySet.csv")

---

End of Session

---