- Dataset: NBA 2023 Player Statistics
- Problem Definition: Analyze the performance metrics of NBA players for the 2023 season.
- Goals:
- Identify top players based on per-game stats
- Find relationships between stats such as shooting efficiency, scoring, assists, and rebounds
- Predict future points per game based on current stats and age
Introduction
Data Overview
## 'data.frame': 539 obs. of 30 variables: ## $ PName: chr "Jayson Tatum" "Joel Embiid" "Luka Doncic" "Shai Gilgeous-Alexander" ... ## $ POS : chr "SF" "C" "PG" "PG" ... ## $ Team : chr "BOS" "PHI" "DAL" "OKC" ... ## $ Age : int 25 29 24 24 28 21 28 26 24 28 ... ## $ GP : int 74 66 66 68 63 79 77 68 73 77 ... ## $ W : int 52 43 33 33 47 40 44 44 38 38 ... ## $ L : int 22 23 33 35 16 39 33 24 35 39 ... ## $ Min : num 2732 2284 2390 2416 2024 ... ## $ PTS : int 2225 2183 2138 2135 1959 1946 1936 1922 1914 1913 ... ## $ FGM : int 727 728 719 704 707 707 658 679 597 673 ... ## $ FGA : int 1559 1328 1449 1381 1278 1541 1432 1402 1390 1388 ... ## $ FG. : num 46.6 54.8 49.6 51 55.3 45.9 45.9 48.4 42.9 48.5 ... ## $ X3PM : int 240 66 185 58 47 213 218 245 154 204 ... ## $ X3PA : int 686 200 541 168 171 578 636 635 460 544 ... ## $ X3P. : num 35 33 34.2 34.5 27.5 36.9 34.3 38.6 33.5 37.5 ... ## $ FTM : int 531 661 515 669 498 319 402 319 566 363 ... ## $ FTA : int 622 771 694 739 772 422 531 368 639 428 ... ## $ FT. : num 85.4 85.7 74.2 90.5 64.5 75.6 75.7 86.7 88.6 84.8 ... ## $ OREB : int 78 113 54 59 137 47 141 63 56 42 ... ## $ DREB : int 571 557 515 270 605 411 626 226 161 303 ... ## $ REB : int 649 670 569 329 742 458 767 289 217 345 ... ## $ AST : int 342 274 529 371 359 350 316 301 741 327 ... ## $ TOV : int 213 226 236 192 246 259 216 180 300 194 ... ## $ STL : int 78 66 90 112 52 125 49 99 80 69 ... ## $ BLK : int 51 112 33 65 51 58 21 27 9 18 ... ## $ PF : int 160 205 166 192 197 186 233 168 104 159 ... ## $ FP : int 3691 3706 3747 3425 3451 3311 3324 2918 3253 2885 ... ## $ DD2 : int 31 39 36 3 46 9 40 5 40 2 ... ## $ TD3 : int 1 1 10 0 6 0 0 0 0 0 ... ## $ X... : int 470 424 128 149 341 97 170 338 100 18 ...
- The dataset contains key statistics like Points (PTS), Assists (AST), Rebounds (REB), Field Goal Percentage (FG%), and more.
Data Cleaning
# libraries library(dplyr)
## ## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats': ## ## filter, lag
## The following objects are masked from 'package:base': ## ## intersect, setdiff, setequal, union
library(caret)
## Loading required package: ggplot2
## Loading required package: lattice
missing_values <- colSums(is.na(nba_data)) missing_values[missing_values > 0]
## named numeric(0)
- No significant missing values found in the dataset, so we proceed with exploratory data analysis.
Exploratory Data Analysis
Per-Game Summary Stats - Code
# per-game stats
nba_data <- nba_data %>%
mutate(PPG = PTS / GP,
RPG = REB / GP,
APG = AST / GP,
TOPG = TOV / GP,
SPG = STL / GP,
BPG = BLK / GP)
colnames(nba_data) <- make.names(colnames(nba_data))
Per-Game Summary Statistics
## PPG RPG APG TOPG ## Min. : 0.000 Min. : 0.000 Min. : 0.0000 Min. :0.0000 ## 1st Qu.: 4.153 1st Qu.: 1.831 1st Qu.: 0.7871 1st Qu.:0.5185 ## Median : 7.033 Median : 3.033 Median : 1.3704 Median :0.8800 ## Mean : 9.119 Mean : 3.541 Mean : 2.0681 Mean :1.1039 ## 3rd Qu.:12.092 3rd Qu.: 4.521 3rd Qu.: 2.7448 3rd Qu.:1.5000 ## Max. :33.076 Max. :12.536 Max. :10.6552 Max. :4.1096 ## SPG BPG ## Min. :0.0000 Min. :0.0000 ## 1st Qu.:0.3211 1st Qu.:0.1389 ## Median :0.5571 Median :0.2727 ## Mean :0.6075 Mean :0.3801 ## 3rd Qu.:0.8333 3rd Qu.:0.4968 ## Max. :3.0000 Max. :3.0000
- Summary statistics for per-game metrics allow players to be compared on average stats to mitigate disparities caused by missing games.
Summary Statistics for Key Metrics
# Summary stats
key_metrics <- c("PTS", "AST", "REB", "FG.", "X3P.", "FT.", "TOV", "STL", "BLK", "PF")
nba_statistics <- summary(nba_data[, key_metrics])
print(nba_statistics)
## PTS AST REB FG. ## Min. : 0.0 Min. : 0.0 Min. : 0.0 Min. : 0.00 ## 1st Qu.: 120.5 1st Qu.: 22.0 1st Qu.: 50.5 1st Qu.: 41.65 ## Median : 374.0 Median : 69.0 Median :159.0 Median : 45.50 ## Mean : 523.4 Mean :115.5 Mean :198.3 Mean : 46.33 ## 3rd Qu.: 769.5 3rd Qu.:162.5 3rd Qu.:286.0 3rd Qu.: 50.60 ## Max. :2225.0 Max. :741.0 Max. :973.0 Max. :100.00 ## X3P. FT. TOV STL ## Min. : 0.00 Min. : 0.00 Min. : 0.0 Min. : 0.00 ## 1st Qu.: 28.10 1st Qu.: 66.70 1st Qu.: 14.5 1st Qu.: 8.50 ## Median : 34.20 Median : 76.30 Median : 44.0 Median : 28.00 ## Mean : 31.53 Mean : 71.99 Mean : 61.3 Mean : 33.27 ## 3rd Qu.: 38.50 3rd Qu.: 84.10 3rd Qu.: 92.5 3rd Qu.: 51.00 ## Max. :100.00 Max. :100.00 Max. :300.0 Max. :128.00 ## BLK PF ## Min. : 0.00 Min. : 0.00 ## 1st Qu.: 5.00 1st Qu.: 32.00 ## Median : 13.00 Median : 86.00 ## Mean : 21.24 Mean : 91.18 ## 3rd Qu.: 28.00 3rd Qu.:140.00 ## Max. :193.00 Max. :279.00
- Descriptive statistics give us an overview of the player metrics, including mean, median, and ranges.
Questions
- Who is the best player based on overall stats?
- Which teams had the highest average points per game among all players?
- What is the relationship between a player’s position and PPG?
- How do FG% and 3PT% affect a player’s scoring?
- How does a player’s age affect performance metrics?
- Can a player’s future PPG be predicted based on current stats and age?
Best Player Analysis (Based on Efficiency)
# efficiency formula nba_data <- nba_data %>% mutate(Efficiency = (PTS + REB + AST + STL + BLK) - ((FGA - FGM) + TOV))
## PName Team Efficiency PPG APG RPG FG. X3P. FT. ## 1 Nikola Jokic DEN 2696 24.49275 9.826087 11.84058 63.2 38.3 82.2 ## 2 Domantas Sabonis SAC 2569 19.11392 7.253165 12.31646 61.5 37.3 74.2 ## 3 Joel Embiid PHI 2479 33.07576 4.151515 10.15152 54.8 33.0 85.7
- The top 3 players are identified using a custom efficiency metric, which considers scoring, assists, rebounds, steals, blocks, missed shots, and turnovers.
Team-wise Performance
# team avg ppg team_avg_pts <- nba_data %>% group_by(Team) %>% summarize(avg_ppg = mean(PPG, na.rm = TRUE)) %>% arrange(desc(avg_ppg)) %>% head(5) print(team_avg_pts)
## # A tibble: 5 × 2 ## Team avg_ppg ## <chr> <dbl> ## 1 PHX 10.8 ## 2 NOP 10.5 ## 3 PHI 10.4 ## 4 LAC 10.4 ## 5 LAL 9.92
- The Phoenix Suns had the highest average points per player, which shows that their ball distribution is among the league’s best.
Player Position vs Points Scored
- The box plot shows that point guards had the highest median PPG, while centers typically scored the least. However, the league leader in PPG was a center, which shows as an outlier.
Field Goal Percentage vs Points Scored
- It can be seen in the graph that players who average at least 20 PPG tend to shoot between 40% - 60%, which is a range that is considered elite in the NBA.
3PT Percentage vs Points Scored - Graph
3PT Percentage vs. Points Scored - Reasoning
- Similarly, players who average at least 20 PPG shoot between 30% - 45% from the three-point line, which suggests a slightly more variable ability to shoot among these scorers.
Age vs. Performance Graphs
Age vs. Performance
- On all of the previous graphs, it is clear that as age increases, PPG/APG/RPG all tend to decrease.
- However, the league’s oldest player, LeBron James, appears as an outlier on all of the graphs.
Predicting Points Per Game Based on Age
# library
library(caret)
# ppg
nba_data <- nba_data %>%
mutate(Above_Current_PPG = ifelse(PPG > median(PPG, na.rm = TRUE), 'Above', 'Below'))
set.seed(123)
nba_data_ml <- nba_data %>%
select(Above_Current_PPG, Age)
data_ml <- na.omit(nba_data_ml)
trainIndex <- createDataPartition(data_ml$Above_Current_PPG, p = .8, list = FALSE)
trainData <- data_ml[trainIndex,]
testData <- data_ml[-trainIndex,]
# logistic trainer
model_logistic <- train(Above_Current_PPG ~ Age, data = trainData, method = "glm", family = "binomial")
# test
predictions <- predict(model_logistic, newdata = testData)
predictions <- factor(predictions, levels = c("Above", "Below"))
testData$Above_Current_PPG <- factor(testData$Above_Current_PPG, levels = c("Above", "Below"))
confusionMatrix(predictions, testData$Above_Current_PPG)
## Confusion Matrix and Statistics ## ## Reference ## Prediction Above Below ## Above 26 22 ## Below 27 32 ## ## Accuracy : 0.5421 ## 95% CI : (0.443, 0.6388) ## No Information Rate : 0.5047 ## P-Value [Acc > NIR] : 0.2494 ## ## Kappa : 0.0832 ## ## Mcnemar's Test P-Value : 0.5677 ## ## Sensitivity : 0.4906 ## Specificity : 0.5926 ## Pos Pred Value : 0.5417 ## Neg Pred Value : 0.5424 ## Prevalence : 0.4953 ## Detection Rate : 0.2430 ## Detection Prevalence : 0.4486 ## Balanced Accuracy : 0.5416 ## ## 'Positive' Class : Above ##
Player predictions
## PName Age PPG Predicted_Above_Current_PPG ## 1 Joel Embiid 29 33.07576 Above ## 2 Shai Gilgeous-Alexander 24 31.39706 Below ## 3 Nikola Jokic 28 24.49275 Above ## 4 Jalen Brunson 26 24.01471 Above ## 5 LeBron James 38 28.90909 Above ## Predicted_PPG_Avg ## 1 36.38333 ## 2 28.25735 ## 3 26.94203 ## 4 26.41618 ## 5 31.80000
Conclusion
- The best player was identified based on a custom efficiency metric, which takes into account scoring, assists, rebounds, steals, blocks, missed field goals, and turnovers, providing measurable impact of a player.
- Analyzing a player’s position and scoring can help create lineups and offensive sets for teams.
- Field goal percentage and three-point percentage were found to correlate with a player’s scoring ability, which demonstrates the effect of shooting percentages on PPG.
- By looking at a player’s age compared to their averages, teams can effectively manage contracts for older players, while the regression can predict a player’s future scoring.