ShadyStats

1
ShadyStats

• Final Report
• Mike Cora
• December 19, 2005
• 533C Information Visualization

2
Background and Motivation

• Spent the last year developing the high-level AI
  for Super Mario Strikers (decisions and
  positioning)
• Coding, tuning and testing a video games AI
  involved gathering a lot of multidimensional
  statistics
• Task goals for a visualization solution
• Show trends, and correlations, compare teams and
  difficulty levels
• Help with finding bugs during development
• Help with tuning a well balanced game
• Aid communication with the publisher (across
  potential language barriers for example)

3
Proposal ShadyStats
2) Hierarchical parallel coordinates
1) Dataset history
4 Dynamic filtering
3 Zoomable, detailed graphs
4
Feature Goals

• Maintain history of datasets for easy comparison
  during development lifecycle
• Use XmdvTool1 for parallel coordinates
• Hierarchical aggregation, with fancy shading
• Extend with interactive drag reordering of
  dimensions
• Generate detailed graphs on demand, using
  Zedgraph2, by interacting with the parallel
  coordinates component

5
Feature Goals 2

• In depth, dynamic filtering and highlighting
  options
• Various statistical measures (mean, variance)
• Highly interactive, tooltip information
  everywhere (data values, dimension names, etc).
• Contextual zooming of individual graphs

6
Demo Parallel Coordinates

• Automatic PCA clustering not applicable, needed
  direct control over clustering.
• Hiding, reordering dimensions. Should be direct
  interaction with parallel coordinates control,
  rather than clunky button interface.
• Add Side cluster, compare by dataset, difficulty,
  side, all data.
• Highlight dataset, difficulty, side hierarchy
  zoom.
• Zooming, panning of parallel coordinates.
• Highlite PassReceive and Possession time
  anomalies.

7
Demo Parallel Coordinates
8
Demo Graphs

• Add Shots vs. Goals
• Add Fouls vs. Goals vs. Sides and Difficulty
  (overlaying curves).
• Magnify and delete graphs, naïve, quick and dirty
  layout, possibly use Piccolo
• Highlight apparent trends.
• Highlight live data linking between all views.
• Built-in zooming, panning and saving features of
  ZedGraph.

9
Demo Graphs
10
Demo Filtering

• Add filter SideHome, DifficultyEasy
• Highly customizable and/or filter tree.
• Everything is linked and updated dynamically.
• More complex filters dirty, high-scoring games,
  home/away difficulty comparison.

11
Demo Filtering
12
Demo Data views

• All view parameters stored in one structure
• Ability to save and switch between arbitrary
  number of views
• Very easy to implement undo/redo system
• Data view copy stored for each view manipulation

13
Unfinished/Future Work

• Fix bugs!!
• Optimize!! (dont use doubles, fix possible
  OpenGL - C bottleneck).
• Undo/Redo system based on save data views
• More interactivity with the parallel coordinate
  control (dragging, selecting, highlighting)
• Data tooltips everywhere
• Coloring of parallel coordinates more meaningful,
  rather than automatic
• Link colors between parallel coordinates and
  graphs more.

14
Unfinished/Future Work

• Improve mean line, more statistical analysis
• Highlighting of specific games by selecting data
  points in parallel coord or graphs.
• Improve filtering interface, may be a bit
  cumbersome
• Possibly use Piccolo for graph layout
• Evaluation
• After the hollidays, our QA Lead will evaluate
  the ShadyStats, apply it to our datasets, and
  suggest features and improvements.

15
Bibliography

• XmdvTool http//davis.wpi.edu/xmdv/
• Zedgraph http//zedgraph.sourceforge.net/
• Edward J. Wegman. Hyperdimensional Data Analysis
  Using Parallel Coordinates, Journal of the
  American Statistical Association, Vol. 85, No.
  411. (Sep., 1990), pp. 664-675.
• Ying-Huey Fua, Matthew O. Ward, and Elke A.
  Rundensteiner, Hierarchical Parallel Coordinates
  for Visualizing Large Multivariate Data Sets,
  IEEE Visualization '99.
• Jing Yang, Wei Peng, Matthew O. Ward and Elke A.
  Rundensteiner, Interactive Hierarchical Dimension
  Ordering, Spacing and Exploration of High
  Dimensional Datasets, Proc. InfoVis 2003.
• John K. Ousterhout, Tcl and the Tk Toolkit,
  Addison Wesley,1994.