Network Viz 4

go to Network Viz 1
go to Network Viz 2
go to Network Viz 3

brightness tests:


number of lights = 1

number of lights = 2

number of lights = 3

number of lights = 4

number of lights = 5

number of lights = 6

number of lights = 7


weapons used by players -- 10,840 players averaged into 43 groups of 256 each -- plot shows that weapon #2 (Weapon_M249_SAW) and weapon #10 (M4A1_Rifle_Mod) are used vastly more often than all others, which have relatively neglible levels
17-Aug-2004


same data as block plot ("citscape") showing individual values more clearly
17-Aug-2004

algorithm test of moving points apart:


iterations=0

iterations=1

iterations=2

iterations=3

iterations=4

iterations=5
18-Aug-2004


18-Aug-2004


visualizing stats from CMU: X=mission type (left-right), Y=player number (up-down), Z=weapon type (in-out of screen); countours surround constant values, animation ranging from 1 to 10 (i.e., everything outside a closed curve is less than the current constant)
23-Aug-2004


visualizing stats from CMU: X=mission type (left-right), Z=weapon type (in-out of screen), Time=player number; now the values are logs of actual stats (stats range from 0 to 65K, logs from 0 to 5); in this GIF animation time goes from 0 to 19, but in on-screen display full dataset is used, going from 0 to 70K
23-Aug-2004


matrix of player number vs. player number: who shot who, based on damage events from one player to another, including self-inflicted frag and fall damage, total size 5080x5080, averaged down to fit in 635x635 pixel image
24-Aug-2004


matrix of player number vs. player number: who shot who, based on damage events from one player to another, including self-inflicted frag and fall damage, total size 5080x5080, close-up of upper left 635x635 pixel region
24-Aug-2004



data from mission: data/input/ClientData6-19/games/logs/64.74.138.156/04-06-18_01-08-11.stats, showing activites of 6 of the 17 players, animated over time displaying the following events:

Each event is a ray from the center to the vertex of a distorted icosahedron (20 faces, 12 corners). If there were no events in a time periood there is a a short ray, and numbers of events are normalized so that the max is always a long ray of the same length. This technique looks very useful for spotting correlations in behavior.
26-Aug-2004



close-up of an icosahedron, showing the 12 ray structure with transparent icosahedron skin
26-Aug-2004


this is the most complex AVS network built for this project so far
26-Aug-2004

Last update  9:52 PM Thu. 26-Aug-2004 by ABS.