NBA Defense Evaluation Using Machine Learning

 
MACHINE LEARNING
EXAMINATION OF NBA
DEFENSE
 
Alex Block
 
Advisors: Chris Fernandes & Nick Webb
 
Project Goal
 
Determine the attributes that are most characteristic of
defensive effectiveness
 
Dean Oliver
 
S
p
e
c
i
f
i
c
a
l
l
y
,
 
a
t
 
t
h
e
 
N
B
A
 
l
e
v
e
l
,
 
t
h
e
 
g
e
n
e
r
a
l
 
r
a
n
k
i
n
g
 
o
f
i
m
p
o
r
t
a
n
c
e
 
o
f
 
t
h
e
 
F
o
u
r
 
F
a
c
t
o
r
s
 
i
s
 
1.
Shooting percentage (40%)
2.
Turnovers per possession (25%)
3.
Offensive rebounding percentage (20%)
4.
Getting to the foul line (15%)
 
 
 
 
Kirk Goldsberry
 
Process Outline
Gather Data
Attribute Selection
Model Data
Analyze
 
Data
 
Data gathered for all teams from the 1996-1997 to the
present season
 
65 Initial Statistics
 
 
Gather Data
Attribute Selection
Model data
Analyze
 
Measuring Defense
 
Defensive Efficiency
Points normalized by 100 possessions
 
 
 
 
 
 
 
 
Gather Data
Attribute Selection
Model Data
Analyze
 
Machine Learning
 
Weka
Apply every available algorithm to the data
 
 
 
 
 
 
Best Performing Algorithms
Linear Regression
Multilayer Perceptron
Gather complete set of
data
Attribute Selection
Model data with
Machine Learning
Analyze
 
Linear Regression
 
y = a + b
1
x
1
 + b
2
x
2
 … b
n
x
n
 
 
 
 
 
 
 
 
 
 
 
 
 
Correlation Coefficient: 0.9997
 
 
 
 
 
 
 
 
 
 
 
 
 
Gather data
Attribute Selection
Model Data
Analyze
 
Multilayer Perceptron (Neural Network)
Gather Data
Attribute Selection
Model Data
Analyze
 
 
 
 
 
 
 
 
 
 
 
 
Correlation Coefficient: 0.9993
 
Attribute Selection
 
Attribute is…
…not an indicator of defense
…an indicator of defensive effectiveness, but not a cause of it
…highly correlated with another attribute
…a composite of other attributes
…not important for the success of the model
Gather Data
Attribute Selection
Model Data
Analyze
 
Final Attribute Set (Ranked by Importance)
 
1.
Opponent Turnover Ratio
2.
Restricted Area Field Goal %
3.
Defensive Rebound %
4.
Mid-Range Field Goal %
5.
Opponent Free Throw Attempt Ratio
6.
In The Paint (Non-RA) Field Goal %
7.
Above the Break 3 Field Goal %
8.
In The Paint (Non-RA) Field Goal Attempts
9.
Mid-Range Field Goal Attempts
10.
Restricted Area Field Goal Attempts
11.
Opponent Free Throw %
12.
Left Corner 3 Field Goal %
13.
Right Corner 3 Field Goal %
14.
Above the Break 3 Field Goal Attempts
15.
Right Corner 3 Field Goal Attempts
16.
Left Corner 3 Field Goal Attempts
 
My Statistics
 
1.
Opp TO Ratio
2.
Restricted Area FG Percentage
3.
DREB Percentage
4.
Mid-Range FG Percentage
5.
Opp FTA Rate
6.
In The Paint (Non-RA) FG Percentage
7.
Above the Break 3 FG Percentage
8.
In The Paint (Non-RA) FGA
9.
Mid-Range FGA
10.
Restricted Area FGA
11.
Opp FT Percentage
12.
Left Corner 3 FG Percentage
13.
Right Corner 3 FG Percentage
14.
Above the Break 3 FGA
15.
Right Corner 3 FGA
16.
Left Corner 3 FGA
 
1.
Shooting percentage
2.
Turnovers per possession
3.
Offensive rebounding percentage
4.
Getting to the foul line
 
Dean Oliver’s
 
My Statistics
 
1.
Opp TO Ratio
2.
Restricted Area FG Percentage
3.
DREB Percentage
4.
Mid-Range FG Percentage
5.
O
p
p
 
F
T
A
 
R
a
t
e
6.
In The Paint (Non-RA) FG Percentage
7.
Above the Break 3 FG Percentage
8.
In The Paint (Non-RA) FGA
9.
Mid-Range FGA
10.
Restricted Area FGA
11.
Opp FT Percentage
12.
Left Corner 3 FG Percentage
13.
Right Corner 3 FG Percentage
14.
Above the Break 3 FGA
15.
Right Corner 3 FGA
16.
Left Corner 3 FGA
 
1.
Shooting percentage
2.
Turnovers per possession
3.
Offensive rebounding percentage
4.
G
e
t
t
i
n
g
 
t
o
 
t
h
e
 
f
o
u
l
 
l
i
n
e
 
Dean Oliver’s
 
My Statistics
 
1.
Opp TO Ratio
2.
Restricted Area FG Percentage
3.
D
R
E
B
 
P
e
r
c
e
n
t
a
g
e
4.
Mid-Range FG Percentage
5.
Opp FTA Rate
6.
In The Paint (Non-RA) FG Percentage
7.
Above the Break 3 FG Percentage
8.
In The Paint (Non-RA) FGA
9.
Mid-Range FGA
10.
Restricted Area FGA
11.
Opp FT Percentage
12.
Left Corner 3 FG Percentage
13.
Right Corner 3 FG Percentage
14.
Above the Break 3 FGA
15.
Right Corner 3 FGA
16.
Left Corner 3 FGA
 
1.
Shooting percentage
2.
Turnovers per possession
3.
O
f
f
e
n
s
i
v
e
 
r
e
b
o
u
n
d
i
n
g
 
p
e
r
c
e
n
t
a
g
e
4.
Getting to the foul line
 
Dean Oliver’s
 
My Statistics
 
1.
O
p
p
 
T
O
 
R
a
t
i
o
2.
Restricted Area FG Percentage
3.
DREB Percentage
4.
Mid-Range FG Percentage
5.
Opp FTA Rate
6.
In The Paint (Non-RA) FG Percentage
7.
Above the Break 3 FG Percentage
8.
In The Paint (Non-RA) FGA
9.
Mid-Range FGA
10.
Restricted Area FGA
11.
Opp FT Percentage
12.
Left Corner 3 FG Percentage
13.
Right Corner 3 FG Percentage
14.
Above the Break 3 FGA
15.
Right Corner 3 FGA
16.
Left Corner 3 FGA
 
1.
Shooting percentage
2.
T
u
r
n
o
v
e
r
s
 
p
e
r
 
p
o
s
s
e
s
s
i
o
n
3.
Offensive rebounding percentage
4.
Getting to the foul line
 
Dean Oliver’s
 
My Statistics
 
1.
Opp TO Ratio
2.
R
e
s
t
r
i
c
t
e
d
 
A
r
e
a
 
F
G
 
P
e
r
c
e
n
t
a
g
e
3.
DREB Percentage
4.
M
i
d
-
R
a
n
g
e
 
F
G
 
P
e
r
c
e
n
t
a
g
e
5.
Opp FTA Rate
6.
I
n
 
T
h
e
 
P
a
i
n
t
 
(
N
o
n
-
R
A
)
 
F
G
 
P
e
r
c
e
n
t
a
g
e
7.
A
b
o
v
e
 
t
h
e
 
B
r
e
a
k
 
3
 
F
G
 
P
e
r
c
e
n
t
a
g
e
8.
I
n
 
T
h
e
 
P
a
i
n
t
 
(
N
o
n
-
R
A
)
 
F
G
A
9.
M
i
d
-
R
a
n
g
e
 
F
G
A
10.
R
e
s
t
r
i
c
t
e
d
 
A
r
e
a
 
F
G
A
11.
O
p
p
 
F
T
 
P
e
r
c
e
n
t
a
g
e
12.
L
e
f
t
 
C
o
r
n
e
r
 
3
 
F
G
 
P
e
r
c
e
n
t
a
g
e
13.
R
i
g
h
t
 
C
o
r
n
e
r
 
3
 
F
G
 
P
e
r
c
e
n
t
a
g
e
14.
A
b
o
v
e
 
t
h
e
 
B
r
e
a
k
 
3
 
F
G
A
15.
R
i
g
h
t
 
C
o
r
n
e
r
 
3
 
F
G
A
16.
L
e
f
t
 
C
o
r
n
e
r
 
3
 
F
G
A
 
1.
S
h
o
o
t
i
n
g
 
p
e
r
c
e
n
t
a
g
e
2.
Turnovers per possession
3.
Offensive rebounding percentage
4.
Getting to the foul line
 
Dean Oliver’s
 
 
Attribute Weights
 
Removed each attribute individually and tracked the
change in the Correlation Coefficient
 
Summed the individual changes in Correlation
 
Divided the individual change by the total change and
multiplied by 100
Weights are a percentage of 100
 
Attribute Weights
 
Attribute Weights
Gather Data
Attribute Selection
Model Data
Analyze
 
Attempts vs. Percentage
 
Brad Stevens (Boston Celtics Coach) on the difference
between college and NBA defense
In college, we really focus on who shoots the ball. In the NBA, there
is much more of a focus on where they shoot from.
 
 
Attempts vs. Percentage
 
Attempts vs. Percentage
 
The Opponent’s FG Percentage is more important than
the Attempts for 
every
 zone
 
Explanation:
Little variation in attempts compared to the variations in percentage
 
Confirmation:
 
Important Zones to Defend
 
Restricted Area
Mid-Range
In The Paint (Non-RA)
Above the Break 3
Corner 3
 
 
These are the differentiators between good/bad defense
Not necessarily the ranked list of most important areas to defend
 
Future work…
 
Attribute Weights by Year
Has the 3 Pointer become more important?
 
Integrating players and player statistics
 
Evaluating offense using the same process
Slide Note

Im _______ my advisors are _____

Embed
Share

Explore the attributes influencing NBA defensive effectiveness through a machine learning examination conducted by Alex Block Advisors, Chris Fernandes, and Nick Webb. The study analyzes various factors such as shooting percentage, turnovers, offensive rebounding, and free throws. Data from the 1996-1997 season to the present is used to measure defensive efficiency and apply machine learning algorithms like Linear Regression and Multilayer Perceptron. Attribute selection plays a crucial role in determining defensive effectiveness.

  • NBA
  • Defense
  • Machine Learning
  • Attribute Selection
  • Data Analysis

Uploaded on Sep 14, 2024 | 0 Views


Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

E N D

Presentation Transcript


  1. MACHINE LEARNING EXAMINATION OF NBA DEFENSE Alex Block Advisors: Chris Fernandes & Nick Webb

  2. Project Goal Determine the attributes that are most characteristic of defensive effectiveness

  3. Dean Oliver Specifically, at the NBA level, the general ranking of importance of the Four Factors is 1. Shooting percentage (40%) 2. Turnovers per possession (25%) 3. Offensive rebounding percentage (20%) 4. Getting to the foul line (15%)

  4. Kirk Goldsberry

  5. Process Outline Gather Data Attribute Selection Model Data Analyze

  6. Data Data gathered for all teams from the 1996-1997 to the present season 65 Initial Statistics Gather Data Attribute Selection Model data Analyze

  7. Measuring Defense Defensive Efficiency Points normalized by 100 possessions Gather Data Attribute Selection Model Data Analyze

  8. Machine Learning Weka Apply every available algorithm to the data Gather complete set of data Best Performing Algorithms Linear Regression Multilayer Perceptron Model data with Machine Learning Attribute Selection Analyze

  9. Linear Regression y = a + b1x1 + b2x2 bnxn Gather data Attribute Selection Model Data Correlation Coefficient: 0.9997 Analyze

  10. Multilayer Perceptron (Neural Network) Gather Data Attribute Selection Model Data Correlation Coefficient: 0.9993 Analyze

  11. Attribute Selection Attribute is not an indicator of defense an indicator of defensive effectiveness, but not a cause of it highly correlated with another attribute a composite of other attributes not important for the success of the model Gather Data Attribute Selection Model Data Analyze

  12. Final Attribute Set (Ranked by Importance) Correlation Coefficient Initial Opponent Turnover Ratio Restricted Area Field Goal % Defensive Rebound % Mid-Range Field Goal % Opponent Free Throw Attempt Ratio In The Paint (Non-RA) Field Goal % Above the Break 3 Field Goal % In The Paint (Non-RA) Field Goal Attempts Mid-Range Field Goal Attempts 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Restricted Area Field Goal Attempts 11. Opponent Free Throw % 12. Left Corner 3 Field Goal % 13. Right Corner 3 Field Goal % 14. Above the Break 3 Field Goal Attempts 15. Right Corner 3 Field Goal Attempts 16. Left Corner 3 Field Goal Attempts Final Multilayer Perceptron Linear Regression .9993 .9952 .9997 .9816

  13. My Statistics Dean Oliver s Opp TO Ratio Restricted Area FG Percentage DREB Percentage Mid-Range FG Percentage Opp FTA Rate In The Paint (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Paint (Non-RA) FGA Mid-Range FGA Restricted Area FGA Opp FT Percentage Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 1. 2. 3. 4. Shooting percentage Turnovers per possession Offensive rebounding percentage Getting to the foul line

  14. My Statistics Dean Oliver s Opp TO Ratio Restricted Area FG Percentage DREB Percentage Mid-Range FG Percentage Opp FTA Rate In The Paint (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Paint (Non-RA) FGA Mid-Range FGA Restricted Area FGA Opp FT Percentage Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 1. 2. 3. 4. Shooting percentage Turnovers per possession Offensive rebounding percentage Getting to the foul line

  15. My Statistics Dean Oliver s Opp TO Ratio Restricted Area FG Percentage DREB Percentage Mid-Range FG Percentage Opp FTA Rate In The Paint (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Paint (Non-RA) FGA Mid-Range FGA Restricted Area FGA Opp FT Percentage Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 1. 2. 3. 4. Shooting percentage Turnovers per possession Offensive rebounding percentage Getting to the foul line

  16. My Statistics Dean Oliver s Opp TO Ratio Restricted Area FG Percentage DREB Percentage Mid-Range FG Percentage Opp FTA Rate In The Paint (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Paint (Non-RA) FGA Mid-Range FGA Restricted Area FGA Opp FT Percentage Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 1. 2. 3. 4. Shooting percentage Turnovers per possession Offensive rebounding percentage Getting to the foul line

  17. My Statistics Dean Oliver s Opp TO Ratio Restricted Area FG Percentage DREB Percentage Mid-Range FG Percentage Opp FTA Rate In The Paint (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Paint (Non-RA) FGA Mid-Range FGA Restricted Area FGA Opp FT Percentage Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 1. 2. 3. 4. Shooting percentage Turnovers per possession Offensive rebounding percentage Getting to the foul line

  18. Attribute Weights Removed each attribute individually and tracked the change in the Correlation Coefficient Summed the individual changes in Correlation Divided the individual change by the total change and multiplied by 100 Weights are a percentage of 100

  19. Attribute Weights Opp TO Ratio Restricted Area FG Percentage DREB Percentage Mid-Range FG Percentage Opp FTA Rate In The Pain (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Pain (Non-RA) FGA Mid-Range FGA Restricted Area FGA Opp FT Percentage Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 27.55453502 25.83237658 13.37543054 8.811710677 6.601607348 4.592422503 4.47761194 1.607347876 1.463834673 1.377726751 1.205510907 0.975889782 0.918484501 0.861079219 0.229621125 0.114810563

  20. Attribute Weights Dean Oliver Alex Block 40% 25% Shooting Percentage Turnovers Per Possession Offensive Rebounding Percentage Getting to the Foul Line 53.4% 27.6% 20% 13.4% 15% 6.6% Gather Data Attribute Selection Model Data Analyze

  21. Attempts vs. Percentage Brad Stevens (Boston Celtics Coach) on the difference between college and NBA defense In college, we really focus on who shoots the ball. In the NBA, there is much more of a focus on where they shoot from.

  22. Attempts vs. Percentage Restricted Area FG Percentage Mid-Range FG Percentage In The Paint (Non-RA) FG Percentage Above the Break 3 FG Percentage In The Pain (Non-RA) FGA Mid-Range FGA Restricted Area FGA Left Corner 3 FG Percentage Right Corner 3 FG Percentage Above the Break 3 FGA Right Corner 3 FGA Left Corner 3 FGA 25.83237658 8.811710677 4.592422503 4.47761194 1.607347876 1.463834673 1.377726751 0.975889782 0.918484501 0.861079219 0.229621125 0.114810563

  23. Attempts vs. Percentage The Opponent s FG Percentage is more important than the Attempts for every zone Explanation: Little variation in attempts compared to the variations in percentage Confirmation: Correlation Coefficient Removing all FGA Removing all FG% 0.9301 0.7826

  24. Important Zones to Defend Restricted Area Mid-Range In The Paint (Non-RA) Above the Break 3 Corner 3 These are the differentiators between good/bad defense Not necessarily the ranked list of most important areas to defend

  25. Future work Attribute Weights by Year Has the 3 Pointer become more important? Integrating players and player statistics Evaluating offense using the same process

Related


More Related Content

giItT1WQy@!-/#giItT1WQy@!-/#giItT1WQy@!-/#