top of page

Beyond Positions: Redefining Player Roles in Modern Basketball

Editor's Note: Player stat glossary can be found at the bottom of the article

It's 2024. The game of basketball has evolved into an age of beautifully spaced, dynamic offense that can confound even historically great defenses. This is a far cry from the game Dr. Naismith first created, a time where positions were defined during a version of basketball where players were not even allowed to dribble. So, it's time to let go.

The traditional labels of "guard," "forward," and "center" no longer capture what a player actually does on the court. Consider Stephen Curry and Ben Simmons: both are labeled as point guards, yet their roles are vastly different. Watch Curry for just two minutes as he dazzles defenders with his supreme shooting and off-ball movement, and it's clear he has a fundamentally different job from Simmons, whose impact is felt primarily within the free-throw line.

As teams evolve and strategies shift, we need to move beyond these outdated categories and redefine player roles based on their true responsibilities. Welcome to the present day of basketball analytics, where "Playmaking Shot Creators," "Roll and Cut Bigs," and "Snipers" take center stage, offering a more accurate and insightful view of how each player shapes the game.

In this article, we'll explore the offensive and defensive roles I've created, examine year-by-year trends, and review the methods used in this approach to understanding basketball.



Playmaking Shot Creator

NBA Comp: James Harden, 2018

ORL Comp: Paolo Banchero, 2024

A perimeter player who carries a heavy offensive load for their team. These players excel as ball handlers and have a high volume of shot-creation (Isolation, Pull Up), playmaking (PnR Ball Handler, Potential Assists) and rim attack (Drives, FTA). Rarely given stationary shooting opportunities (Corner 3PA, Catch and Shoot), and don’t often complete off-ball actions.

Key Stats:

  • Isolation

  • PnR Ball Handler

  • Pull Up

  • Potential Assists

  • Drives

  • FTA

Traditional Point

NBA Comp: Tyus Jones, 2024

ORL Comp: D.J. Augustin, 2019

A player who primarily handles the ball and creates opportunities for their teammates. They have high volume in traditional ball handling actions (Pick and Roll Ball Handler, Drives, Potential Assists) and focus less at trying to create their own shot (Isolation, Pullup, Handoff).

Key Stats:

  • PnR Ball Handler

  • Potential Assists

  • Drives


NBA Comp: Duncan Robinson, 2024

ORL Comp: Terrence Ross, 2019

A player who generates almost all their offensive volume on 3-point shooting opportunities. The offenses they play on generally create schemes to get them Catch and Shoot opportunities. They take lots of shot attempts off Screens and Handoffs, and have an above average volume of Pull Up jump shots.

Key Stats:

  • 3PA

  • Handoff

  • Off Screen

  • Catch and Shoot


Corner Spacer

NBA Comp: Matisse Thybulle, 2024

ORL Comp: Chuma Okeke, 2024

A low volume player whose offensive volume is almost exclusively through stationary shooting opportunities (Spot Up and Catch and Shoot). These players take most of their shots from the corner. They rarely get chances through Handoff and Off Screen actions.

Key Stats:

  • Corner 3PA

  • Catch and Shoot

  • Spot Up

  • 3PA


Swiss Army Knife

NBA Comp: Christian Braun, 2024

ORL Comp: Anthony Black, 2024

A player who creates opportunities on offense in all facets without having one sole specialty. Typically, a wing that does not rely on shooting or high-volume ball handling to generate offense. These players crash the glass, attack the rim and cut at middle frequencies. When they shoot it will often be a spot up shot from the corner.

Key Stats:

  • Rim FGA

  • Cut

  • Offensive Rebound

  • Corner 3PA


Interior Playmaker

NBA Comp: Nikola Jokic, 2024

ORL Comp: N/A

A player who creates playmaking and scoring opportunities at and around the rim. High volume of Post Ups, Isolation, PnR Roll, and FTA. They not only create opportunities for themselves to score but create a relatively high number of potential assists for their teammates.

Key Stats:

  • Rim FGA

  • Post Up

  • Potential Assists

  • FTA

  • PnR Roll

  • Isolation


Roll and Cut Big

NBA Comp: Daniel Gafford, 2024

ORL Comp: Goga Bitadze, 2024

A big whose offensive volume is almost entirely generated through rim running actions (Rim FGA, PnR Roll, Cut). These players are the highest volume glass crashers in the league, reaching high percentiles for ORebs, Boxouts and Putbacks. Their only self-generated offense is through Post Ups.

Key Stats:

  • Rim FGA

  • PnR Roll

  • Cut

  • OReb


Versatile Big

NBA Comp: Nikola Vucevic, 2024

ORL Comp: Wendell Carter Jr., 2024

A big who generates offense in many ways. They operate as a PnR Roller at a high volume, can create their own scoring from the post as well as create spacing as a shooter. These players are less adept at passing (than an Interior Playmaker), and offensive rebounding (than a Roll and Cut big).

Key Stats:

  • Post Up

  • PnR Roll

  • Catch & Shoot

  • OReb



Point of Attack

NBA Comp: Jaden McDaniels, 2024

ORL Comp: Jalen Suggs, 2024

A defender who matches up against the opposition's strongest initiator, generally a guard. They are the first point of contact for the defense who aims to disrupt offensive actions before they start. High volume in defending PnR Ball Handlers, Handoffs and Off Screen possessions. Very high matchup difficulty.

Key Stats:

  • Matchup Difficulty

  • PnR Ball Handler

  • Handoff

  • Off Screen

  • Deflections

  • vs Guard


NBA Comp: Devonte Graham, 2024

ORL Comp: Joe Ingles, 2024

A player whose defense aims to keep away from defending any action. They have extremely low matchup difficulty, and are smaller players, mostly facing up against guards (to not be exploited for size). Offenses may try to attack them by bringing their matchup into Off Screen, Handoff and PnR Ball Handler situations. They contest shots at a low rate.

Key Stats:

  • Matchup Difficulty

  • PnR Ball Handler

  • Handoff

  • Off Screen

  • Height

  • vs Guard

Secondary Guard

NBA Comp: Devin Vassell, 2024

ORL Comp: Cole Anthony, 2024

A defensive guard who matches up on the opponent's second-option guard. Defenses are not trying to hide these players, but they avoid matching them up on the offense's best players. As a result, they are often chasing off-guards around screens and handoffs.

Key stats:

  • PnR Ball Handler

  • Handoff

  • Off Screen

Lockdown Wing

NBA Comp: Kawhi Leonard, 2024

ORL Comp: Aaron Gordon, 2020

A defensive player who matches up against the opponent's strongest wings, particularly forwards. A player that is typically a very versatile defender, who has to face all types of offensive actions. They have solid defensive volume in every perimeter and interior category.

Key stats:

  • Matchup Difficulty

  • vs Forward

  • Perimeter Defense (all categories)

  • Interior Defense (all categories)

Secondary Forward

NBA Comp: Julius Randle, 2024

ORL Comp: Paolo Banchero, 2024

A defensive wing who matches up on the opponent's second-best forward. They don't have to be hidden as they are generally taller players who cannot be easily exploited on the interior. However, they have generally slower footspeed, so defenses avoid matching them up with quick guards. These players most commonly defend Corner 3s and Isolations, but also have some volume of interior defense and can contribute to rebounding.

Key stats:

  • vs Forward

  • Matchup Difficulty

  • Corner 3PA

  • Isolation

Switchable Big

NBA Comp: Evan Mobley, 2024

ORL Comp: Mo Wagner, 2023

A big who defends a high volume of interior actions, but who can also be swtiched onto wings. These players are often tasked with defending large playmakers and will face a high volume of Isolations as well as Contested 3s. They combine decent footwork with some interior presence.

Key stats:

  • vs Center

  • vs Forward

  • Isolation

  • Contested 3s

  • Interior Defense (All Categories)

Paint Anchor

NBA Comp: Rudy Gobert, 2024

ORL Comp: Nikola Vucevic, 2018

A big who sets themself up at the ring, and rarely defends other areas of the court. Their primary role is as a paint protector, and as the player who faces opposition Post Ups. These players are often the tallest players in the league and are less mobile, meaning that defenses avoid switching them onto wings. Generally, they face an opposition's strongest center. They are pivotal in cleaning the glass.

Key stats:

  • vs Center

  • Height

  • Interior Defense (All Categories)



Observing player roles between 2018 and 2024 allows us to understand how NBA teams adjust the roles of their players to adjust to their competition. Here, I will break down how many total playing minutes there were for each role in the NBA in these seasons.

(Credit to Hua Gong and Su Chen who presented this analysis for their roles between 2015 and 2022 at the MIT Sloan Sports Analytics Conference).

Offensive Roles between 2018 and 2024

A few key movers within the NBA's offensive roles over the last 7 years.

  1. Corner Spacers are the most prominent role in 2024. They have increased their proportion by 5% over the observed period. These players provide bail out spacing for drives.

  2. Traditional Points have decreased from being the most common role to the third. This move has coincided with an increase in Playmaking Shot Creators and Interior Playmakers, who can generate team offense without the need for a Traditional Point.

  3. Interior Playmakers are coming back. They are still the least common offensive role, but teams have been finding more minutes for these players over the last few years.

Defensive Roles

In the past 7 years, the defensive landscape has shifted as offenses have unlocked the 3-point line and the spacing that it provides (thank you Golden State and Houston). With this, players are often forced into more and more actions, as perimeter players have dominated on offense.

This comes with:

  • A drop in the number of Secondary Guards, who are forced into a more active on ball role.

  • An increase in the number of Lockdown Wings, as teams are giving more minutes to versatile players who can deal with the nightmare wings who can score 30 points on any given night.

  • A slow decrease in Paint Anchors over the past 7 years, as these players have been frequently targeted in the Playoffs. It will be interesting to see if this shifts in coming years, with dominant bigs like Jokic and Embiid being key playoff pieces.



It would be very reasonable to ask yourself how players are even assigned roles. Is it just a someone with a lot of spare time giving players different names? Or is it possible for someone with a lot of spare time to actually use player statistics to create the roles for them?

This is where I introduce Role Clustering - the process that makes this all possible. Simply put, role clustering is the process of taking the stats of every player in the NBA and finding groups within that data. At the simplest level, take this scatter plot with a player's 2 Point Attempts (2PA) on the X-Axis, and their 3 Point Attempts (3PA) on the Y-Axis. Here, you might see 4 groups, or clusters, appear:

  1. Players who attempt lots of 3s and 2s

  2. Players who only attempt lots of 3s

  3. Players who only attempt lots of 2s

  4. Players who don't attempt many 3s or 2s

Voila, you have Role Clustering. The only difference being - you use 20+ stats instead of 2.

To do this there are 4 key steps:

  1. Choose the stats that you want to include.

  2. Choose the number of clusters you want to find.

  3. Find clusters / groups of players using your chosen stats.

  4. Train a model to assign new players to these clusters.

Choosing the stats for your clusters

Choosing exactly what to include is the most important part of role clustering. You need to clearly understand what facets of the game you want to capture. If you want to find a player's role, the number one thing to consider is to use only play frequency data. This means that you group players based on what they actually do, not by how good they are. Otherwise, your clusters end up becoming good players and bad players. This is not the point of the exercise. You want to understand what a player's job is.

It is also important to use per minute stats. Similar to avoiding stats that describe how good a player is, stats at the per game or total level will result in clusters of players that do or don't play a lot of minutes. This would not tell you what a player actually does.

Thankfully, role clustering work has been done before by some awesome data analysts such as Todd Whitehead (Synergy), Xu Lian, BBall Index, Paola Zuccolotto, Marica Manisera, and many more. By reviewing some of their work, I had a baseline to adjust and figure out exactly what I was after.

Offense Target Areas

For offense I landed on:

  1. On-Ball Creation

  2. System Shooting

  3. Stationary Shooting

  4. Rim Attack

  5. Glass Crash

Defense Target Areas

For defense, the categories ended up being more straightforward:

  1. Perimeter Defense

  2. Interior Defense

  3. Matchups

Choosing the number of clusters

Choosing the number of groups for cluster analysis is your next key consideration. You must balance the tradeoff between how precise you want your clusters to be, and how easily interpretable they are. Having more clusters can capture finer details in the data and improve the accuracy and granularity of each group, but it also increases the risk of overfitting and can make interpretation harder. The extreme example here would be that you could have perfectly accurate clusters by creating one for every player in the NBA, and call them 'the Paolo Banchero cluster', 'the Franz Wagner cluster' etc. However, the results of this would be useless.

On the other hand, fewer clusters simplify the model and make it easier to understand and explain, but having too few clusters may overlook subtle patterns and important distinctions within the data. Striking the right balance is important to ensure that the clusters are both meaningful and manageable.

So how can you make this decision? One common approach is to use the elbow method. Here, you can plot how the variance of each group (how different the players within each group are) decreases with each number of clusters. You want to find a point that removes as much variance as possible without creating too many groups. Here, you can see the elbow charts for Offense and Defense:

You can see that the variance is close to minimised around the 7 or 8 cluster mark. This, combined with some other methods provide you with a ballpark number of clusters for each side of the ball.

The reality of choosing a final number of clusters is that it ends up becoming a painstaking process of trial and error. You try a number of clusters in this range and ask yourself - "Do these groups make sense? Are there any groups missing?" When the answer to both questions is yes, you have your clusters.

Finding groups within these clusters

You now have a lot of stats, a lot of players, and a lot of groups. So how do you go about organizing players into clusters? Principal Component Analysis (PCA). The process of PCA clustering is like organizing a messy garage. First, you look for the most important stuff (stats that become Principal Components) and group similar items together, like all your tools in one corner and sports gear in another. Then, you use a method like k-means clustering to put these groups into boxes, making them easy to find and understand what’s in your garage.

Here, for role clustering, you can find groups of similar players that allow us to quickly identify what a player's role on the court might be. There's a lot going on, but you can see the general groups of data that this process finds in the charts below:

Offensive Clusters

Defensive Clusters

Training a model to assign new players to clusters

The final step is to create a way to assign new players to clusters. You do this so that you don't have to do this process all over again every time that there are new players or seasons (Important to note that a player's role can change between seasons). This is done through machine learning. Very simply, you take 90% of the players used for clustering (withholding 10% as a control group that allows you to test the output of your model) and tell a computer to look at all of their stats and learn what a player in each cluster looks like. The model this creates gives you the probability that a player is in each cluster. The cluster with the highest probability then becomes the player's role. A player whose second most likely cluster is > 30% chance, has this role become their "Secondary Role".

But you're not quite done. Before you do a celebratory lap of your living room, you need to know if your model is predicting roles correctly. To do this, you test it with the 10% of players that you did not show your computer. When the computer manages to predict each role with > 90% success, you can have some level of confidence that your model make sense.

Congratulations, you officially have player roles. Now, it's time to write an unneccessarily long article about them.


Player Stat Glossary

Cluster Stat



3 Point Attempts


Block - A defensive player tipping a ball that has been shot, blocking the chance to score


A player making physical contact with an opponent who was actively pursuing a rebound. The opponent must be successfully prevented from securing the rebound

Catch & Shoot

A shot attempted made after catching the ball and making 0 dribbles

Contested 2s

A defensive player closing out and raising a hand to contest a 2 point shot prior to its release

Contested 3s

A defensive player closing out and raising a hand to contest a 3 point shot prior to its release

Corner 3PA

3 Point Attempts from the corner of the court


A movement by a player without the ball towards the basket to receive the ball

Defensive Rebound

A rebound collected while on defense


A defensive player getting their hand on the ball on a non-shot attempt


A player attacking the basket off the dribble in the halfcourt offense. Does not include situations where a player starts close to the basket, or immediately gets cut off on the perimeter


Free Throw Attempts


A player is handed the ball directly from a teammate


Player height


A play designed for a player to take their defender one on one

Matchup Difficulty

The average strength of opponent that a defender is matched up again (measured as their opponent's season points per minute)

Off Screen

A player receives a pass after using a teammates screen to get open

Offensive Rebound

A rebounds collected while on offense

PnR Ball Handler

The player handling the ball in a pick and roll play

PnR Roll

The roll action performed by the screener in a pick and roll play

Post Up

A play where an offensive player backs down a defender near the basket

Potential Assists

A pass to a teammate who shoots within 1 dribble of receiving the ball

Pull Up

A shot taken by a player after immediately after dribbling

Put Back

An attempt to score immediately after grabbing an offensive rebounding


Field Goal Attempts made at the basket

Spot Up

A shot attempt where the player is stationary before attempting the shot

vs Center

Portion of defensive minutes matched up vs centers

vs Forward

Portion of defensive minutes matched up vs forwards

vs Guard

Portion of defensive minutes matched up vs guards

526 views0 comments

Recent Posts

See All


bottom of page