Context
The new era of the NBA requires versatility to win:
“I don’t have the five positions anymore. It may be as simple as three positions now, where you’re either a ball-handler, a wing or a big. It’s really important. We’ve become more versatile as the years have gone on.” Celtics coach Brad Stevens, per Kareem Copeland of the Associated Press.
Even within a position, players can serve significantly different roles. The Warriors have inspired ‘small ball’, emphasizing movement and quick transition offense. On the other side of the spectrum, teams like Denver are building around Nikola Jokic, a ‘big’ listed at 7’0 and 250 pounds. This seems to harken back to a bygone era of the need for a dominant big man, but the comparison gets murky with his ability to shoot the three, average a double-double while ranking in the top 10 in assist percentage.
Advanced metrics have evolved along with the NBA. This project uses ‘RAPTOR’ statistics for example, which were developed last year by FiveThirtyEight Sports and is described as:
“A statistic that better reflects how modern NBA teams actually evaluate players. NBA teams highly value floor spacing, defense and shot creation, and they place relatively little value on traditional big-man skills. RAPTOR likewise values these things — not because we made any deliberate attempt to design the system that way but because the importance of those skills emerges naturally from the data. RAPTOR thinks ball-dominant players such as James Harden and Steph Curry are phenomenally good. It highly values two-way wings such as Kawhi Leonard and Paul George. It can have a love-hate relationship with centers, who are sometimes overvalued in other statistical systems. But it appreciates modern centers such as Nikola Jokić and Joel Embiid, as well as defensive stalwarts like Rudy Gobert.” (Full Article: Introducing RAPTOR, Our New Metric For The Modern NBA )
Prior to the 2018–19 season, the NBA G League announced a Select Contract as part of a comprehensive professional path that will be available, beginning with the 2019–20 season, to elite prospects who are not yet eligible for the NBA. The contracts, which will include robust programmatic opportunities for development, are for elite players who are at least 18 years old and will pay $125,000 for the five-month season.
With the new opportunity to sign players out of high school, strategies to evaluate player potential and development would be valuable to General Managers. What is the best way to maximize ROI in this new era? How quickly should you pull the plug if there are negative trends? What unique, measurable traits can predict a successful player type?
Goals
- Use advanced metrics to better classify modern NBA player roles.
- Using this new lens, develop a machine learning model to predict a young player’s potential NBA classification or player role as they develop.
With this tool, scouts and front office executives would have a better understanding of a young player’s potential career trajectory, track that player’s development progress year over year, and help make decisions on contract commitments.
Data
In order to classify modern nba players, this project used advanced metrics that take advantage of player tracking and play-by-play data that isn’t available in traditional box scores.
- NBA Advanced Metrics from FiveThirtyEight Sports
repository link
We’ll use historical college statistics of our newly classified modern nba players in order to create the prediction model. Cross Validation will be used to train and test the model.
- NCAA College statistics are available from basketball-reference.com.
Results
1. Positionality Exploration Across Eras a. Build b. Evaluate
90’s Era
Modern Era
2. Clustering
Unsupervised clustering was used to generate 6 modern player types. Below you’ll see the Silhouette Score peak at 3 and 6 clusters which was the basis of this determination. While 3 has the highest peak I wanted to be able to differentiate the positions into more than 3 categories.
The graph above shows two of the three principle components due to the 2 dimensional space we are viewing them. The three components comprise 74% of the total variation. The overlapping clusters illustrate that whatever sets those clusters apart isn’t a part of the two principal components used as the axes for the graph. In order to understand what sets the clusters apart we’ll need to look into the players and feature importance within each cluster.
3. Label Identification and Understanding
C0: Two-way long distance shooters, generally not the first scoring option, but accurate from deep.
- Ray Allen
- Danny Green
- JJ Redick
- Kyle Korver
C1: Perimeter role player, 3pt spot up shooters, defensive liability
C2: Close to the rim, defensive minded bigs. No outside shot, but can get to the free throw line. C5 Backup.
- Kendrick Perkins
- Ben Wallace
C3: Versatile sharp-shooters, generally a higher % of shots come from long-distance. C4 potential but can be lacking on defense. Can create offense for themselves and teammates. Poor man’s C4.
- Kyle Lowry
- Jamal Crawford
- Deron Williams
- Tony Parker
C4: No weakness, 2-way players, create offense for themselves and teammates. 1st option, max contract players. Build your team around superstars
- Steph Curry
- Lebron
- Nikola Jokic
- Kevin Durant
C5: Traditional center mold, or PF without range, but a knack for scoring around the rim. Rebounders. Likely your team’s most impactful defender.
- Dwight Howard
- Rudy Gobert
- Tyson Chandler
4. Prediction Model
For the prediction model I ran three logistic regression iterations to predict whether or not a player would be a superstar based on their college statistics. Below are the results. The balanced logistic regression model performed the best.
Scouts and front office executives could potentially use a tool like this for the following purposes:
- View young player’s potential career trajectory
- Track player development year over year, adjust strategies
- Evaluate young players to make decisions on contract commitments
- Better focus scouting on players who would potentially fill needs
