Discovering Meaning in Baseball Statistics Through Information Visualization

We humans are visual creatures. Thus, any tool that might improve our capacity to understand large datasets through visual display of information patterns deserves our attention. After defining and describing infovis (“information visualization”) systems and their principles, Andy Cox presents applications he has developed, using Retrosheet data and other large-scale datasets, to demonstrate the utility of infovis for rapid summarizing, analyzing, filtering, and understanding baseball data and their interrelationships.

Transcript

1. Discovering Meaning in
   Baseball Statistics Through
   Information Visualization
   Andy Cox
   SABR 36
   July 1, 2006
   Friday, September 20, 13
   

   View Slide

2. Copyright © 2006 Andy Cox or by original copyright holders.
   What is
   information visualization?
   Friday, September 20, 13
   

   View Slide

3. Copyright © 2006 Andy Cox or by original copyright holders.
   Information visualization
   Friday, September 20, 13
   

   View Slide

4. Copyright © 2006 Andy Cox or by original copyright holders.
   Information visualization
   • Not the “silver bullet”
   • Another (valuable) tool for the toolbox
   Friday, September 20, 13
   

   View Slide

5. Copyright © 2006 Andy Cox or by original copyright holders.
   Information visualization
   • Not the “silver bullet”
   • Another (valuable) tool for the toolbox
   • No single definition, but general consensus
   • 20+ years as research topic
   Friday, September 20, 13
   

   View Slide

6. Copyright © 2006 Andy Cox or by original copyright holders.
   Information visualization
   • Not the “silver bullet”
   • Another (valuable) tool for the toolbox
   • No single definition, but general consensus
   • 20+ years as research topic
   • “Infovis in under 10 minutes”
   • This is my 5-10 minute summary
   • Nowhere near enough time
   • Plenty of references at the end
   Friday, September 20, 13
   

   View Slide

7. Copyright © 2006 Andy Cox or by original copyright holders.
   Information visualization
   My (simple) working definition:
   Friday, September 20, 13
   

   View Slide

8. Copyright © 2006 Andy Cox or by original copyright holders.
   Information visualization
   Information visualization (infovis) involves
   creating a visual representation of abstract
   information and allowing manipulation of this
   representation to facilitate exploration and
   insight.
   My (simple) working definition:
   Friday, September 20, 13
   

   View Slide

9. Copyright © 2006 Andy Cox or by original copyright holders.
   Huh?
   Friday, September 20, 13
   

   View Slide

10. Copyright © 2006 Andy Cox or by original copyright holders.
    Information visualization
    Information visualization (infovis) involves
    creating a visual representation of abstract
    information and allowing manipulation of this
    representation to facilitate exploration and
    insight.
    My (simple) working definition:
    Friday, September 20, 13
    

    View Slide

11. Copyright © 2006 Andy Cox or by original copyright holders.
    Information visualization
    Information visualization (infovis) involves
    creating a visual representation of abstract
    information and allowing manipulation of this
    representation to facilitate exploration and
    insight.
    My (simple) working definition:
    Friday, September 20, 13
    

    View Slide

12. Copyright © 2006 Andy Cox or by original copyright holders.
    Information visualization
    Information visualization (infovis) involves
    creating a visual representation of abstract
    information and allowing manipulation of this
    representation to facilitate exploration and
    insight.
    My (simple) working definition:
    Friday, September 20, 13
    

    View Slide

13. Copyright © 2006 Andy Cox or by original copyright holders.
    So What?
    Friday, September 20, 13
    

    View Slide

14. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    Friday, September 20, 13
    

    View Slide

15. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    • Useful for finding meaning behind data
    • Nothing new in baseball performance
    analysis
    Friday, September 20, 13
    

    View Slide

16. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    • Useful for finding meaning behind data
    • Nothing new in baseball performance
    analysis
    • Plenty of techniques, tools
    • e.g., linear regression, Excel, R
    Friday, September 20, 13
    

    View Slide

17. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    Friday, September 20, 13
    

    View Slide

18. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    • Often must know specific question
    • Is there a relationship between X and Y?
    Friday, September 20, 13
    

    View Slide

19. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    • Often must know specific question
    • Is there a relationship between X and Y?
    • What if we don’t have a specific question?
    Friday, September 20, 13
    

    View Slide

20. Copyright © 2006 Andy Cox or by original copyright holders.
    Statistical analysis
    • Often must know specific question
    • Is there a relationship between X and Y?
    • What if we don’t have a specific question?
    • Hypothesis testing vs. exploratory data analysis
    (Shneiderman, 2001)
    Friday, September 20, 13
    

    View Slide

21. Copyright © 2006 Andy Cox or by original copyright holders.
    Information visualization
    Information visualization (infovis) involves
    creating a visual representation of abstract
    information and allowing manipulation of this
    representation to facilitate exploration and
    insight.
    My (simple) working definition:
    Friday, September 20, 13
    

    View Slide

22. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    Friday, September 20, 13
    

    View Slide

23. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • Traditional statistical/data graphics
    • Charts and graphs
    • Usage dates back to 1800’s (Playfair)
    • For 2D, hard to beat
    Friday, September 20, 13
    

    View Slide

24. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • Traditional statistical/data graphics
    • Charts and graphs
    • Usage dates back to 1800’s (Playfair)
    • For 2D, hard to beat
    • Better than aggregates
    Friday, September 20, 13
    

    View Slide

25. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • Traditional statistical/data graphics
    • Charts and graphs
    • Usage dates back to 1800’s (Playfair)
    • For 2D, hard to beat
    • Better than aggregates
    • Excel, R, etc. can generate graphics
    Friday, September 20, 13
    

    View Slide

26. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    But...
    Friday, September 20, 13
    

    View Slide

27. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • What about more than 2 (or 3) variables?
    But...
    Friday, September 20, 13
    

    View Slide

28. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • What about more than 2 (or 3) variables?
    • What about more complex data?
    • Hierarchical, networks
    But...
    Friday, September 20, 13
    

    View Slide

29. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    Friday, September 20, 13
    

    View Slide

30. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • Edward Tufte
    • Infovis = Tufte not true!
    Friday, September 20, 13
    

    View Slide

31. Copyright © 2006 Andy Cox or by original copyright holders.
    Visual representation
    • Edward Tufte
    • Infovis = Tufte not true!
    • Principles for designing good graphics
    • Maximize the data/ink ratio
    • Avoid “chartjunk”
    Friday, September 20, 13
    

    View Slide

32. Copyright © 2006 Andy Cox or by original copyright holders.
    Perception
    A simple example
    0932475091273023857109871
    8123739803245798198236731
    7527187391877987791610828
    9735901870398382095809004
    Friday, September 20, 13
    

    View Slide

33. Copyright © 2006 Andy Cox or by original copyright holders.
    Perception
    A simple example
    0932475091273023857109871
    8123739803245798198236731
    7527187391877987791610828
    9735901870398382095809004
    Friday, September 20, 13
    

    View Slide

34. Copyright © 2006 Andy Cox or by original copyright holders.
    Information visualization
    Information visualization (infovis) involves
    creating a visual representation of abstract
    information and allowing manipulation of this
    representation to facilitate exploration and
    insight.
    My (simple) working definition:
    Friday, September 20, 13
    

    View Slide

35. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Friday, September 20, 13
    

    View Slide

36. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Biggest difference between infovis and
    traditional statistical graphics
    Friday, September 20, 13
    

    View Slide

37. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Biggest difference between infovis and
    traditional statistical graphics
    • Facilitates rapid browsing and discovery
    • Select what and how to display
    • (remember hypothesis testing vs.
    exploratory data analysis)
    Friday, September 20, 13
    

    View Slide

38. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Biggest difference between infovis and
    traditional statistical graphics
    • Facilitates rapid browsing and discovery
    • Select what and how to display
    • (remember hypothesis testing vs.
    exploratory data analysis)
    • Maybe not so much answering questions as
    asking better ones...
    Friday, September 20, 13
    

    View Slide

39. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Friday, September 20, 13
    

    View Slide

40. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Selecting data you want to see
    Friday, September 20, 13
    

    View Slide

41. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Selecting data you want to see
    • Selecting a different representation
    Friday, September 20, 13
    

    View Slide

42. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Selecting data you want to see
    • Selecting a different representation
    • Selecting the perspective
    Friday, September 20, 13
    

    View Slide

43. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Example technique: Dynamic Queries
    Friday, September 20, 13
    

    View Slide

44. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Rapid, incremental filtering of data
    Example technique: Dynamic Queries
    Friday, September 20, 13
    

    View Slide

45. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Rapid, incremental filtering of data
    • Immediate and continuous display of results
    Example technique: Dynamic Queries
    Friday, September 20, 13
    

    View Slide

46. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    • Rapid, incremental filtering of data
    • Immediate and continuous display of results
    • Based on idea that questions are imprecise,
    evolve during discovery
    Example technique: Dynamic Queries
    Friday, September 20, 13
    

    View Slide

47. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Example technique: Dynamic Queries
    Source: Maryland HCIL
    Friday, September 20, 13
    

    View Slide

48. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Example technique: Dynamic Queries
    Source: Maryland HCIL
    Friday, September 20, 13
    

    View Slide

49. Copyright © 2006 Andy Cox or by original copyright holders.
    Exploration and insight
    Friday, September 20, 13
    

    View Slide

50. Copyright © 2006 Andy Cox or by original copyright holders.
    Exploration and insight
    Visual Information Seeking Mantra:
    Friday, September 20, 13
    

    View Slide

51. Copyright © 2006 Andy Cox or by original copyright holders.
    Exploration and insight
    Visual Information Seeking Mantra:
    Overview first,
    Friday, September 20, 13
    

    View Slide

52. Copyright © 2006 Andy Cox or by original copyright holders.
    Exploration and insight
    Visual Information Seeking Mantra:
    Overview first,
    zoom and filter,
    Friday, September 20, 13
    

    View Slide

53. Copyright © 2006 Andy Cox or by original copyright holders.
    Exploration and insight
    Visual Information Seeking Mantra:
    Overview first,
    zoom and filter,
    then details-on-demand
    Friday, September 20, 13
    

    View Slide

54. Copyright © 2006 Andy Cox or by original copyright holders.
    Exploration and insight
    Visual Information Seeking Mantra:
    Overview first,
    zoom and filter,
    then details-on-demand
    (Shneiderman, The Eyes Have It, 1996)
    Friday, September 20, 13
    

    View Slide

55. Copyright © 2006 Andy Cox or by original copyright holders.
    Enough talk.
    How about some examples?
    Friday, September 20, 13
    

    View Slide

56. Copyright © 2006 Andy Cox or by original copyright holders.
    Parallel coordinates
    Friday, September 20, 13
    

    View Slide

57. Copyright © 2006 Andy Cox or by original copyright holders.
    Parallel coordinates
    • Helps solve the “more than 2D” problem
    Friday, September 20, 13
    

    View Slide

58. Copyright © 2006 Andy Cox or by original copyright holders.
    Parallel coordinates
    • Helps solve the “more than 2D” problem
    • Detecting relationships in multivariate data
    Friday, September 20, 13
    

    View Slide

59. Copyright © 2006 Andy Cox or by original copyright holders.
    Source: XmdvTool
    Friday, September 20, 13
    

    View Slide

60. Copyright © 2006 Andy Cox or by original copyright holders.
    Map of the Market
    Friday, September 20, 13
    

    View Slide

61. Copyright © 2006 Andy Cox or by original copyright holders.
    Map of the Market
    • SmartMoney.com
    Friday, September 20, 13
    

    View Slide

62. Copyright © 2006 Andy Cox or by original copyright holders.
    Map of the Market
    • SmartMoney.com
    • Treemap technique for stock market data
    • Recursively divides rectangle by some
    attribute of the data
    • Hierarchy: Sector>Industry>Company
    • Area: Market Capitalization ($)
    • Color: Change in stock price
    Friday, September 20, 13
    

    View Slide

63. Copyright © 2006 Andy Cox or by original copyright holders.
    Source: SmartMoney.com
    Friday, September 20, 13
    

    View Slide

64. Copyright © 2006 Andy Cox or by original copyright holders.
    So what does this have to do with baseball?
    Friday, September 20, 13
    

    View Slide

65. Copyright © 2006 Andy Cox or by original copyright holders.
    Baseball examples
    Friday, September 20, 13
    

    View Slide

66. Copyright © 2006 Andy Cox or by original copyright holders.
    Baseball examples
    • Statistical graphics are everywhere
    Friday, September 20, 13
    

    View Slide

67. Copyright © 2006 Andy Cox or by original copyright holders.
    Baseball examples
    • Statistical graphics are everywhere
    • A lot of Excel charts
    Friday, September 20, 13
    

    View Slide

68. Copyright © 2006 Andy Cox or by original copyright holders.
    Baseball examples
    • Statistical graphics are everywhere
    • A lot of Excel charts
    • General infovis techniques and systems that
    use baseball data
    Friday, September 20, 13
    

    View Slide

69. Copyright © 2006 Andy Cox or by original copyright holders.
    Table Lens
    Rao & Card (1994)
    Source: ACM
    Friday, September 20, 13
    

    View Slide

70. Copyright © 2006 Andy Cox or by original copyright holders.
    Sparklines
    • Edward Tufte: “data-intense, design-simple,
    word-sized graphics”
    Beautiful Evidence (2006) Source: Edward Tufte
    Friday, September 20, 13
    

    View Slide

71. Copyright © 2006 Andy Cox or by original copyright holders.
    Sparklines
    • Hardball Times has sparkline generator
    Source: The Hardball Times
    Friday, September 20, 13
    

    View Slide

72. Copyright © 2006 Andy Cox or by original copyright holders.
    Other sports
    Tennis Viewer
    Source: IEEE
    Friday, September 20, 13
    

    View Slide

73. Copyright © 2006 Andy Cox or by original copyright holders.
    But nothing specifically for baseball
    Friday, September 20, 13
    

    View Slide

74. Copyright © 2006 Andy Cox or by original copyright holders.
    Demo
    Friday, September 20, 13
    

    View Slide

75. Copyright © 2006 Andy Cox or by original copyright holders.
    Bar display
    Friday, September 20, 13
    

    View Slide

76. Copyright © 2006 Andy Cox or by original copyright holders.
    Player map
    Friday, September 20, 13
    

    View Slide

77. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Friday, September 20, 13
    

    View Slide

78. Copyright © 2006 Andy Cox or by original copyright holders.
    Interaction
    Friday, September 20, 13
    

    View Slide

79. Copyright © 2006 Andy Cox or by original copyright holders.
    Future ideas
    • Player info per game
    • More granular display
    • Plate appearances vs. games
    • Love to hear yours!
    Friday, September 20, 13
    

    View Slide

80. Copyright © 2006 Andy Cox or by original copyright holders.
    Conclusions
    Friday, September 20, 13
    

    View Slide

81. Copyright © 2006 Andy Cox or by original copyright holders.
    • Not “better than” - “in addition to”
    Conclusions
    Friday, September 20, 13
    

    View Slide

82. Copyright © 2006 Andy Cox or by original copyright holders.
    • Not “better than” - “in addition to”
    • A tool for helping to ask better questions
    Conclusions
    Friday, September 20, 13
    

    View Slide

83. Copyright © 2006 Andy Cox or by original copyright holders.
    • Not “better than” - “in addition to”
    • A tool for helping to ask better questions
    • Think about applications
    Conclusions
    Friday, September 20, 13
    

    View Slide

84. Copyright © 2006 Andy Cox or by original copyright holders.
    References
    • Human-Centered Computing Educational Digital Library
    http://hcc.cc.gatech.edu
    • Readings in Information Visualization, Card, Mackinlay,
    Shneiderman (1998)
    • Information Visualization, Spence (2001)
    • Information Visualization: Perception for Design, Ware (2004)
    • Information Dashboard Design, Few (2006)
    • SmartMoney.com Map of the Market
    http://www.smartmoney.com/maps/
    Friday, September 20, 13
    

    View Slide

85. Copyright © 2006 Andy Cox or by original copyright holders.
    References
    • Ben Shneiderman (University of Maryland)
    • Dynamic queries (1994, with C. Ahlberg)
    • Visual Information Seeking Mantra (1996)
    • Edward Tufte
    • The Visual Display of Quantitative Information (2001 2e)
    • Envisioning Information (1990)
    • Visual Explanations (1997)
    • Beautiful Evidence (2006)
    Friday, September 20, 13
    

    View Slide

86. Copyright © 2006 Andy Cox or by original copyright holders.
    References
    • http://www.sportvis.com
    (My site)
    Friday, September 20, 13
    

    View Slide

87. Copyright © 2006 Andy Cox or by original copyright holders.
    Acknowledgements
    • Retrosheet
    • John Stasko (Georgia Tech)
    • SABR
    Friday, September 20, 13
    

    View Slide

88. Copyright © 2006 Andy Cox or by original copyright holders.
    Thank you!
    http://www.sportvis.com
    andy@sportvis.com
    Friday, September 20, 13
    

    View Slide