CourtVisionPH A System for the Extraction of Field Goal Attempt - - PowerPoint PPT Presentation

▶
courtvisionph
SMART_READER_LITE
LIVE PREVIEW

CourtVisionPH A System for the Extraction of Field Goal Attempt - - PowerPoint PPT Presentation

Jun 23, 2023 312 likes •680 views

CourtVisionPH A System for the Extraction of Field Goal Attempt Locations and Spatial Analysis of Shooting using Broadcast Basketball Videos Engr. Ben Hur S. Pintor Nikko Boy N. Cataniag Asst. Prof. Ma. Rosario Concepcion O. Ang University of

slide-1
SLIDE 1

CourtVisionPH

A System for the Extraction of Field Goal Attempt Locations and Spatial Analysis of Shooting using Broadcast Basketball Videos

FOSS4G Seoul, South Korea | September 14-19, 2015

  • Engr. Ben Hur S. Pintor

Nikko Boy N. Cataniag

  • Asst. Prof. Ma. Rosario Concepcion O. Ang

University of the Philippines Department of Geodetic Engineering

slide-2
SLIDE 2

OUTLINE

  • I. INTRODUCTION
  • II. PROBLEM STATEMENT
  • III. OBJECTIVES
  • IV. METHODOLOGY
  • V. RESULTS AND DISCUSSION
  • VI. CONCLUSIONS AND RECOMMENDATIONS

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-3
SLIDE 3

INTRODUCTION / INSPIRATION

CourtVision -- New Visual and Spatial Analytics for the NBA (Dr. Kirk Goldsberry, MIT Sloan Sports Analytics Conference 2012) Scoring Area = 1248 shooting cells 1ft x 1ft per cell

<images from CourtVision -- New Visual and Spatial Analytics for the NBA (Dr. Kirk Goldsberry, MIT Sloan Sports Analytics Conference 2012)> >

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-4
SLIDE 4

INTRODUCTION / INSPIRATION

CourtVision -- New Visual and Spatial Analytics for the NBA (Dr. Kirk Goldsberry, MIT Sloan Sports Analytics Conference 2012)

n = 1248

Spread - number of unique shooting cells with at least one (1) FGA Spread% - Spread / n Range - number of unique shooting cells with an average of at least 1 Point-per-attempt (1 PPA) Range% - Range / n

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-5
SLIDE 5

INTRODUCTION / INSPIRATION

SportVU Player Tracking System (STATS LLC.)

<image from http://www.tvtechnology.com/news/0002/nba-to-follow-the-bouncing-data/226050>

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-6
SLIDE 6

INTRODUCTION / INSPIRATION

Available Player Position and Tracking Data for the NBA

<image from stats.nba.com/tracking>

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-7
SLIDE 7

INTRODUCTION / INSPIRATION

Basketball is SPATIAL

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-8
SLIDE 8

INTRODUCTION / INSPIRATION

Basketball is #1 Sport in the Philippines

  • History
  • Culture
  • Money

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-9
SLIDE 9

PROBLEM STATEMENT

  • 1. No system in place to gather spatial information from

basketball games

  • 2. Basketball Analysis and Management in the Philippines is

still very Traditional

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-10
SLIDE 10

OBJECTIVES

  • 1. To develop a system that extracts field goal attempt

locations from broadcast basketball videos, performs spatial analysis of shooting, and presents the results of the analysis with statistics and visualizations, using freely available resources.

  • 2. To show the advantages of spatial analysis of shooting
  • ver the conventional non-spatial statistics currently used

in the Philippines.

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-11
SLIDE 11

METHODOLOGY | DEVELOPMENT

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-12
SLIDE 12

METHODOLOGY | DEVELOPMENT

Main Functionalities

  • 1. Data Management
  • 2. Field Goal Attempt Extraction (Manual)
  • 3. Spatial Analysis

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-13
SLIDE 13

METHODOLOGY | DEVELOPMENT

Data Management

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-14
SLIDE 14

METHODOLOGY | DEVELOPMENT

Data Management

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-15
SLIDE 15

METHODOLOGY | DEVELOPMENT

Field Goal Attempt Extraction

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-16
SLIDE 16

METHODOLOGY | DEVELOPMENT

Field Goal Attempt Extraction

2D Projective Coordinate Transformation The scoring area is defined as a 15mx10m grid composed of 1mx1m cells (150 total cells).

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-17
SLIDE 17

METHODOLOGY | DEVELOPMENT

Field Goal Attempt Extraction

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-18
SLIDE 18

METHODOLOGY | DEVELOPMENT

Spatial Analysis

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-19
SLIDE 19

METHODOLOGY | DEVELOPMENT

Spatial Analysis

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-20
SLIDE 20

METHODOLOGY | APPLICATION

After the development of the system, it was used to study the UP Fighting Maroons and the DLSU Green Archers during the 2nd Round of the University Athletics Association of the Philippines (UAAP) Season 76 (2013-2014). Data used:

  • Videos publicly available on youtube.com (Low Quality)
  • Box-scores and play-by-play available online

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-21
SLIDE 21

METHODOLOGY | APPLICATION

Field Goal Attempts excluded from Extraction:

  • FGA outside the scoring area
  • FGA with bad RMSE or back-substitution results

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-22
SLIDE 22

RESULTS AND DISCUSSION

20% (10-12 FGA/game) difference between the number of extracted FGA and those from box-scores.

  • Personal limitation of the user
  • Exclusion of shots beyond the scoring area

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-23
SLIDE 23

RESULTS AND DISCUSSION

Statistics

UP Fightng UP Opponents DLSU Green DLSU Opponents Maroons Archers

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-24
SLIDE 24

RESULTS AND DISCUSSION

Visualizations - Range% (UP, UP Opp)

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-25
SLIDE 25

RESULTS AND DISCUSSION

Visualizations - Range% (DLSU, DLSU Opp)

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-26
SLIDE 26

RESULTS AND DISCUSSION

Visualizations - Range% (Marata - UP, Teng - DLSU)

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-27
SLIDE 27

RESULTS AND DISCUSSION

Observations

  • UP and DLSU have relatively the same distribution of FGA

in terms of distance from the basket

  • UP has a slight advantage over DLSU on shots at the rim

(< 1m from basket) and three-pointers

  • DLSU has a significant advantage on close-range (1m - 3m)

and mid-range (3m - 5m)

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-28
SLIDE 28

RESULTS AND DISCUSSION

Observations

  • UP has difficulties converting and defending high-

percentage shots (< 3m from basket)

  • Only 36% of UP’s FGA were taken within 3m from the

basket at an average rate of 0.99 PPA

  • UP’s opponents took 52% of their shots within 3m from the

basket at a rate of 1.15 PPA

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-29
SLIDE 29

RESULTS AND DISCUSSION

Observations

  • DLSU only allowed their opponents to take 36.61% of their

shots within 3 meters at a rate of 1.10 PPA

  • DLSU was able force their opponents to a lot of mid-range

shots (17.85%) and long 2-pointers (14.15%) where these

  • pponents scored just 0.62 PPA

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-30
SLIDE 30

RESULTS AND DISCUSSION

Observations

  • J. Marata (UP) takes shots from all over the court

(peppered look of his Range% map) but is effective only at certain areas of the court

  • J. Teng’s (DLSU) FGA are concentrated in the paint which

he converts at a high rate, he seldom takes long two- pointers or three-point shots

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-31
SLIDE 31

CONCLUSIONS

A system that can extract field goal attempt locations from broadcast basketball videos and perform spatial analysis of shooting was successfully developed and applied using freely available resources and data. It was demonstrated that spatial analysis provides a better characterization and appreciation of shooting as compared to conventional non-spatial shooting statistics.

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-32
SLIDE 32

CONCLUSIONS

Aside from the quality of the videos used, the system is limited by the ability of the user to completely and correctly extract all the field goal attempts taken during a game.

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-33
SLIDE 33

RECOMMENDATIONS

  • expansion of the database
  • better video and image processing algorithms
  • automatic court detection and tracking
  • automatic shot and position determination
  • use of own cameras/video capture devices or multiple

cameras (i.e. GoPro, off-the-shelf video cameras)

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-34
SLIDE 34

RECOMMENDATIONS

  • data fed into the system must complete and correct
  • use supplementary sources such as play-by-play data and

regular checking and counter-checking of the contents of the database against known values from boxscores

  • allow a person who has intimate knowledge of basketball

to perform the extraction

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH

slide-35
SLIDE 35

REFERENCES

Goldsberry, K. 2012. CourtVision – New Visual and Spatial Analytics for the NBA. MIT Sloan Sports Analytics Conference, Boston. MA, USA, 2-3 March 2012.

FOSS4G Seoul, South Korea | September 14-19, 2015 CourtVisionPH