Redefining Positions and Players In Todays NBA, Using Machine Learning
Using Machine Learning, I turned the 5 old broad positions in basketball to 8 newly defined ones
Introduction
The NBA has changed a lot from when it began in 1950 to what it is today. Some of these changes include rules like the ever famous hand check rule that was first introduced in the 2004–2005 season, teams, such as the Seattle Supersonics becoming the Oklahoma City Thunder, and the different play styles that emerged over the years (for example: emphasis on scoring in the paint to emphasis on the three-pointer). But even though the NBA has changed dramatically with how the game of basketball is played today, the positions that describe these players has remained the same for the past 50 years.
The five Positions that can be used to describe any player in basketball are: 1. Point Guard (PG), 2. Shooting Guard (SG), 3. Small Forward (SF), 4. Power Forward (PF), 5. Center ( C ). Each position has their own role during the game. For example, the Point Guard’s role is to bring the ball up the court and direct plays, while a Center’s role is to be the primary rebounder, contest shots, and set screens on plays. But with the evolution of basketball, players and their assigned positions have become blurred. Many players today have stepped out of the boundary of their respective roles and into another.
Take a look at Lebron James. Lebron’s official position is Small Forward, whose role is to aggressively attack the rim to post up, lay up, and try to draw fouls so they can get to the line. But if you’ve ever watched a game that Lebron was in over the past couple of years, you can see that Lebron plays more of a Point Guard role. He constantly brings the ball up the court to facilitate and directs plays for his team. And Lebron is just one example of the many players in the NBA who merge the role of two, or even three positions at a time in a game.
The current positions that describe players appear to be irrelevant, and new ones should be created to better describe these players’ roles.
Through the use of machine learning, I did just that. Utilizing the data from the 2018–2019 season , I created eight new positions/styles of basketball that can better describe each player and their role.
The Computing Process
1. Data
The data for this project was taken from basketball-reference.com, and contains the stats for each NBA player per game in the 2018–2019 season. The data was than edited to only contain the relevant information required. Therefore, information such as age of the player and the team they played for were removed. The data was also filtered to contain only the players that have played 30 games or more, due to the fact if they played a very low number of games their data might not be entirely accurate or reliable. Finally, the data left was scaled and ready for the next step.
2. Linear Discriminant Analysis (LDA)
Linear Discriminant Analysis is a method used for dimension reduction. It finds a linear combination of features that separates the dependant variable into two or more classes. In simpler terms, it reduces the amount of features, to the ones that best separate the different classes. The purpose for using this, was to reduce the amount of features (independent variables) that impacted the dependant variable.
There are 25 features or independent variables in the dataset, which include FG, FG%, 3P, AST, TRB, PTS, etc. To better find the features that describe each player, Linear Discriminant Analysis was used, with the traditional positions of each player being the dependant variable. This was used to see what features impact the position of each player the most, so than if a new player was inserted into the dataset, it could correctly predict the player’s position based on its features (independent variables or stats).
The result of using this method, was that it reduced the amount of features from 25 to 9 ,and when tested, it correctly predicted a players’ position based on its stats 93% of the time.
(it should be noted that when the features are reduced, they are reduced to different independent variables that the program makes, as seen below)
Using Linear Discriminant Analysis the dataset went from…
To…
3. K-Means Clustering
I then clustered the data into eight clusters using K-Means Clustering.
How K-Means Clustering works is that you define a target number (k), which refers to the number of centroids and clusters you want. A centroid is the imaginary location representing the center of the cluster. A cluster refers to a collection of data points aggregated together because of certain similarities.
Next, every data point is allocated to the nearest cluster while keeping the centroid as small as possible. This means that the variance of the data points to the centroid is quite small.
In the end, eight clusters were created that contained players with certain similarities. These eight clusters will eventually become the different positions/play styles that the players are organized into.
4. Principal Component Analysis (PDA)
Principal Component Analysis is another dimensionality reduction method. But unlike Linear Discriminant Analysis, we can specify how many features we want in the end.
Since the data is organized into eight groups, we can apply the Principal Component Analysis, so that the nine features that describe each group now become two. Two features were chosen so that we can visual on a scatter plot regarding what players belong to what group.
That’s it!!!
We have now created and are able to visualize on a scatterplot, the players that belong in the eight newly defined positions/play styles.
8 Different Positions/Play Styles in Today’s NBA
1. Stationary Bigs
This group of players play more of an old school type of basketball. They rank second in two point field goals and first in total rebounds (includes offensive and defensive rebounds) for, they stay close to the basket and mostly defend and score in the paint. As the name suggests, they don’ t typically move around the floor too much. This leads to them being ranked last in three-pointers, but since they are always near the basket they rank first in blocks and in personal fouls. This is because, as offensive players drive to the basket, they often bump into guys in the paint, which usually results in a foul called against them.
Notable Players
- Anthony Davis
- Joel Embiid
- LaMarcus Aldridge
- Karl-Anthony Towns
- Hassan Whiteside
2. Mobile Bigs
As the counterpart to Stationary Bigs, Mobile Bigs are different in a lot of ways, but also similar. Like Stationary Bigs, Mobile Bigs contain players that are mostly centers and power forwards, and they are very good on the defensive side of the floor. They rank near the top for blocks, defensive rebounds, and steals. What makes the two different is on the offensive side of the floor. This group of players don’t take as much two point field goals as Stationary Bigs, but they make much more three-pointers and are typically better at offense.
Notable Players
- Nikola Jokic
- Pascal Siakam
- Marc Gasol
- Rudy Gay
- Aaron Gordon
3. Designated Scorers
As the title suggests, these players are offensive scoring machines, and contain a mix of players and positions from shooting guards, point guards, and small forwards. These players can score from nearly any location on the floor, have great handles, and are easily able to get to their spot to try for a high-percentage shot. They dominate in almost every single offensive category ranking first in: field goals, two point field goals, three point field goals, free throws made, free throw percentage, and points. They may be the most star studded group, but for a good reason, as they rank first in usage percentage, meaning in their team’s plays they are utilized more than anyone else.
Notable Players
- Stephen Curry
- Klay Thompson
- Devin Booker
- Kemba Walker
- Victor Oladipo
4. Floor Generals
Referred to as the quarterbacks of the NBA, these players are the captains and facilitators of their team when they are on the floor. They are the ones who carry the ball up the court and direct the plays and pace of their team. Like a quarterback in the NFL, if they’re having a bad game, the entire team usually doesn’t play well. As these players help everyone else for, they rank first in assists which mean they are spreading the ball around, letting guys hit their shots. They also rank first in steals, meaning they are giving their offence more opportunities to score a bucket and get points. These players may not have the eye-popping stats like the designated scorers, but they’re a major reason why those players have such outstanding numbers.
Notable Players
- Lebron James
- Kyrie Irving
- Damian Lillard
- Chris Paul
- Russel Westbrook
5. Combo Players
What could be described as the most complete group, Combo Players have been referred to in the NBA as two-way players. These players play at an above-average level on both sides of the floor. They can attack on the offensive and get a bucket, then come back and play lock-down defence. Their rankings display this too; in every offensive and defensive category they rank above average or in the middle. There is really nothing they are exceptional at, like how 3-Point Specialist’s are great at three-pointers, but there is also nothing they are really bad at either. They are just complete, well-rounded basketball players.
Notable Players
- Kawhi Leonard
- Paul George
- James Harden
- Jimmy Butler
- Kyle Kuzma
6. 3-Point Specialist
The three point line has become the emphasis in today’s NBA which has led to an overwhelming about of three-point shots’ being taken compared to a few years ago. That is what gives this group of players importance in today’s NBA. As 3-Point Specialists, they are just that, exceptionally good at getting three-pointers and not much of anything else. As their ranks go, they rank very high in terms of three point field goals, three point field goals attempted, and three point field goal percentage. But when it comes to all the other categories, they are mostly below average. Therefore, having a full team of three point specialists may not be the best idea, but during crunch time they have become very valuable when a three is needed.
Notable Players
- Kyle Korver
- Danny Green
- Seth Curry
- Channing Frye
- Terrence Ross
7. Versatile Forwards
Imagine combining a small forward and a power forward together. That is what this group is comprised of; players that often play the role of both of these positions during a game. It makes sense too, as this group is mostly comprised of small forwards and power forwards. Like a small forward, these players are able to drive to the basket and shoot from long range. Additionally, they can score close to the basket while also being able to shoot mid-range jump shots like a power forward. Offensively and defensively they rank in about the middle, but they excel at getting rebounds for their above- average length, being ranked third in total rebounds just behind both of the big groups.
Notable Players
- Kevin Durant
- Giannis Antetokounmpo
- Tobias Harris
- Andre Iguodala
- Brook Lopez
8. Role Players
Even though their stats may not show it, these players are vital to any team. When looking at their ranks, they rank near the bottom in almost every category, but that doesn’t mean they aren’t any less important. In fact, these players help the stars on their team produce at such a high level. This is because when the stars or starters on the team get tired, role players come off the bench and are put in to the game to give their team quality minutes. They give the stars and starters much needed rest and peace of mind knowing that since they are off the floor, these role players can come in and keep a lead for the team or just compete at a high level for a couple of minutes.
Notable Players
- Terry Rozier
- Brandon Ingram
- Rodney Hood
- Lance Stephenson
- Zaza Pachulia
Further Analysis
Using advanced statistics, we can see how each newly defined position affects each game and how important each new position is.
Using the mean data from each group, the three features that show this and that are going to be analyzed are Player Efficiency Rating, Usage Percentage, and Win Shares.
- Player Efficiency Rating (PER)
Player Efficiency Rating is a per-minute rating of the player. It sums up all of a player’s positive accomplishments, subtracts the negative accomplishments, and returns the per-minute rating of a player’s performance.
It’s not surprising that both big groups are the most efficient. These players don’t really take risks, as they stay close to the basket and take high- percentage shots, thus leading to more positive accomplishments in the game. They also make a positive contribution to their team by ranking near the top in total rebounds, which increases their efficiency rating. On the opposite spectrum, however, 3-Point Specialists, who have become very popular in today’s NBA, are near the bottom only behind the Role Players. Using the same logic as above, this is because their shots are riskier and harder to make. Also since they don’t really get any rebounds, their efficiency rating is almost entirely based on how they shoot in the game.
This leads to a major question in today’s NBA. Would you rather have a team filled with three-point shooters and major scorers like the Golden State Warriors. Where if players like Steph Curry, Klay Thompson, or Kevin Durant shoot poorly, they will have a low chance to win the game, for their efficiency on the court would be really low? Or would you rather have a team filled with big players who can shoot and protect the paint like the Toronto Raptors? In this case, even if players like Kawhi Leonard, Marc Gasol, and Pascal Siakam are shooting bad, they can at least play solid defence and crash the boards to get rebounds, resulting in more offensive possessions.
It should be noted that both of these teams are in the NBA finals right now, as this study is being conducted.
2. Usage Percentage (USG%)
Usage Percentage is an estimate of the percentage of team plays used by a player while he/she was on the floor.
One group stands out in particular, and that’s the Designated Scorers. Every other group has about the same usage percentage, but Designated Scorers have an even greater percentage. This shows that when they’re on the offensive, these players are used the most in their team’s plays. The reason for this is because, what is the objective when a team is on offence? It’s to get the ball in the net, and who’s the best at doing that? It’s the Designated Scorers, who rank first in nearly every offensive category. So it should be no surprise that the players who can go and get a bucket are also the players that are used the most.
3. Win Shares (WS)
Win shares is a player statistic that estimates a number of wins contributed by each player.
In today’s NBA, with the emergence of the three-point line, there has been much talk as to whether these big players were still needed. The Golden State Warriors gave evidence of this, for they have won three of the last four titles with a small team filled with shooters, and not many big players. This starts the conversation as to why other teams don’t just do the same thing. And this may just be the reason, as the groups that have the highest win shares are actually both of the big groups. Demonstrating that even in a high-scoring league filled with great shooters, these big players are still a major part of a team’s success.
Take a look at the team’s with the top records in the NBA this year.
- Milwaukee Bucks (60 wins)
- Toronto Raptors (58 wins)
- Golden State Warriors (57 wins)
- Denver Nuggets (54 wins)
Three of the top four teams are filled with players that are in the big groups, have length, and have good but not great shooting skills like the Golden State Warriors, who in fact is the outlier is this situation.
Conclusion
I made these eight new positions based on the data given in the 2018–2019 season, because I felt there needed to be a change in how players are described. For too long have they been described using the same boring five positions that don’t truly give any explanation of what the player actually does. So through the use machine learning, I turned those five previous broad positions into eight newly defined ones. These new positions are better at describing each player, for factual stats were used to show their strengths/weaknesses, and since there is a greater number of positions, the roles of each position are more explicit.
Even though eight new positions were created, it doesn’t mean that those eight will last forever, or even in 20 years. The fact is that basketball can change in an instant. One player can come in and dominate the game with his or her style and inevitably change how every team plays. That’s exactly what Steph Curry did when he entered the league in 2009. He dominated with his incredible three-point shots, and now every team emphasizes that skill.
The main point of this study is that basketball, and nearly all sports do change. But we can help it change by providing in-depth analytics so that new people coming into the sport can better understand it. My hope is that fans and people just introduced to basketball can see this study and obtain greater knowledge of the game and the players who play it. I also hope that teams see this study and use the results of it to incorporate it into their own play style and determine which combinations of players give any team the best chance to win.
