Interevent time distributions 
of human multi-level activity 
in a virtual world 
Yurij Holovatch 
Lviv, Ukraine
In collaboration with: 
Olesya Mryglod 
Stefan Thurner 
Benedikt Fuchs 
Michael Szell 
Supported by: 
IRSES N612707 (Dyonicos)
MOTIVATION 
² Distribution functions characterising sequences of human actions over time are 
highly non-trivial, and their origins remain largely unclear: writing letters (Oliveira’05, 
Malmgren’09), checking out books, performing financial transactions (V´azquez’05), 
writing e-mails (Kleinberg’03, Bar´abasi’07), web browsing, editing Wikipedia (Jo’12, 
Yasseri’12) etc. 
² The common feature of the above mentioned observations is that they are based 
on the analysis of human actions of a single type: either writing e-mails, or letters, 
or making phone calls, etc. 
² What is the action dynamics of multi-level human activity (writing both e-mails, 
and letters, and making phone calls, etc). 
² A testing ground: virtual world.
MMOG PARDUS 
² Online since 2004 
² Open-ended game 
² Self-organized social 
environment 
² More than 400,000 registered 
players (>14,000 active) 
² economic life (mining, trade, 
production and consumption of 
commodities, etc) 
² social life (chat, forum, private 
messages, establishing friendships 
or hostile relations, status demon-stration)
MULTILEVEL HUMAN ACTIVITY 
Player 146  ... AAAAAACTT   EEX     FTTTTTX CCCTTTTT AC  ... 
Player 199  ... CCA BBCAAAAATTA AACCCCCBX          CFFFF ... 
Player 701  ... CCCCTTTT   TCTCT  FF       CXXTT   CCCCC TTT   ... 
Player 171  ... AAAACC  CCC  C     CCC AA TTT  FCC       EED  ... 
² Communication (C) 
² Trade (T) 
² Marking friends in the list (F) 
² Removing enemies from the list 
(X) 
² Attack (A) 
² Marking enemies in the list (E) 
² Removing friends from the list 
(D)
INTEREVENT TIME DISTRIBUTION 
100 
10−2 
10−4 
10−6 
104 105 106 107 108 10−8 
t [sec] 
Binned P(t) 
−2.09 
(a) 
bin size: 6h 
106 
10−4 
102 103 104 105 
100 
10−2 
10−4 
t [sec] 
Binned P(t) 
−1.12 
(b) 
bin size: 1min 
a. b. 
Distribution of the interevent time ¿ for all players who performed at least 50 
actions. (a) entire observation period (1,238 days), bin size is 6 hours=21; 600 
sec. (b) first 24 hours, bin size is 1 min. Inset: same as (a) for for six days. 
Circadian rhythms are clearly visible.
BURSTINESS 
B = ¾¡m 
¾+m 
(a) 
(b) 
(c) 
(d) 
(e) 
(f) 
B=0.94 
B=-0.38 
B=0.006 
B=0.53 
B=0.53 
B=0.53 
Action streams of players with different values of 
burstiness B. Lines mark times of executed actions, 
the distance between lines is the interevent time. 
Pardus (all action types): B ' 0:53. 
Mobile phone communication (Jo et al., 2012): B ' 0:2. 
Wikipedia editing (Yasseri et al., 2012): B ' 0:6.
ACTION-SPECIFIC DYNAMICS 
Inverse cumulative distribution of interevent 
times ¿ for different kinds of actions. Inset: 
same plot in log-linear scale, for ¿ > 2 ¢ 107 
sec (> 8 months). 
² Immediate reaction (¿ · 360 sec) 
² Early day (6 min < ¿ < 8 hours): 
P¸(¿) » ¿¡® 
² Late day (8 h. < ¿ < 24 h.): 
P¸(¿) » exp (¡¿=¿0) 
² Long times (¿ > 2 ¢ 107 sec)
GLOBAL DYNAMICS AND ACTIVITY PATTERNS: 
WAR AND PEACE 
16000 
14000 
12000 
10000 
8000 
6000 
4000 
0 500 1000 
Time [days] 
Number of actions 
war I war II war III 
Number of actions per day on the time-line. 
The three coloured vertical stripes 
indicate war periods, the vertical line in-dicates 
the introduction of a major new 
game feature. 
Faction #1 
Faction #2 
Faction #3 
Neutral 
200 400 600 800 1000 1200 
3 
2.5 
2 
1.5 
1 
0.5 
x 104 
Time [days] 
Number of communication events 
(a) 
war I war II war III 
Number of weekly actions for different 
factions of players. Activity of players in-volved 
in wars is increased. (a) commu-nication 
C.
CONCLUSIONS 
² Interevent time distributions in the Pardus universe are highly non-trivial. 
² Presence of periodic patterns on different scales. 
² Bursty dynamics. 
² Different kinds of actions are characterized by different decay constants. 
² Periods of increased activity in the Pardus due to history specific events. 
² Physica A 419, 681-690 (2014) [arXiv:1407.2006]

Interevent time distributions of human multi-level activity in a virtual world

  • 1.
    Interevent time distributions of human multi-level activity in a virtual world Yurij Holovatch Lviv, Ukraine
  • 2.
    In collaboration with: Olesya Mryglod Stefan Thurner Benedikt Fuchs Michael Szell Supported by: IRSES N612707 (Dyonicos)
  • 3.
    MOTIVATION ² Distributionfunctions characterising sequences of human actions over time are highly non-trivial, and their origins remain largely unclear: writing letters (Oliveira’05, Malmgren’09), checking out books, performing financial transactions (V´azquez’05), writing e-mails (Kleinberg’03, Bar´abasi’07), web browsing, editing Wikipedia (Jo’12, Yasseri’12) etc. ² The common feature of the above mentioned observations is that they are based on the analysis of human actions of a single type: either writing e-mails, or letters, or making phone calls, etc. ² What is the action dynamics of multi-level human activity (writing both e-mails, and letters, and making phone calls, etc). ² A testing ground: virtual world.
  • 4.
    MMOG PARDUS ²Online since 2004 ² Open-ended game ² Self-organized social environment ² More than 400,000 registered players (>14,000 active) ² economic life (mining, trade, production and consumption of commodities, etc) ² social life (chat, forum, private messages, establishing friendships or hostile relations, status demon-stration)
  • 5.
    MULTILEVEL HUMAN ACTIVITY Player 146 ... AAAAAACTT EEX FTTTTTX CCCTTTTT AC ... Player 199 ... CCA BBCAAAAATTA AACCCCCBX CFFFF ... Player 701 ... CCCCTTTT TCTCT FF CXXTT CCCCC TTT ... Player 171 ... AAAACC CCC C CCC AA TTT FCC EED ... ² Communication (C) ² Trade (T) ² Marking friends in the list (F) ² Removing enemies from the list (X) ² Attack (A) ² Marking enemies in the list (E) ² Removing friends from the list (D)
  • 6.
    INTEREVENT TIME DISTRIBUTION 100 10−2 10−4 10−6 104 105 106 107 108 10−8 t [sec] Binned P(t) −2.09 (a) bin size: 6h 106 10−4 102 103 104 105 100 10−2 10−4 t [sec] Binned P(t) −1.12 (b) bin size: 1min a. b. Distribution of the interevent time ¿ for all players who performed at least 50 actions. (a) entire observation period (1,238 days), bin size is 6 hours=21; 600 sec. (b) first 24 hours, bin size is 1 min. Inset: same as (a) for for six days. Circadian rhythms are clearly visible.
  • 7.
    BURSTINESS B =¾¡m ¾+m (a) (b) (c) (d) (e) (f) B=0.94 B=-0.38 B=0.006 B=0.53 B=0.53 B=0.53 Action streams of players with different values of burstiness B. Lines mark times of executed actions, the distance between lines is the interevent time. Pardus (all action types): B ' 0:53. Mobile phone communication (Jo et al., 2012): B ' 0:2. Wikipedia editing (Yasseri et al., 2012): B ' 0:6.
  • 8.
    ACTION-SPECIFIC DYNAMICS Inversecumulative distribution of interevent times ¿ for different kinds of actions. Inset: same plot in log-linear scale, for ¿ > 2 ¢ 107 sec (> 8 months). ² Immediate reaction (¿ · 360 sec) ² Early day (6 min < ¿ < 8 hours): P¸(¿) » ¿¡® ² Late day (8 h. < ¿ < 24 h.): P¸(¿) » exp (¡¿=¿0) ² Long times (¿ > 2 ¢ 107 sec)
  • 9.
    GLOBAL DYNAMICS ANDACTIVITY PATTERNS: WAR AND PEACE 16000 14000 12000 10000 8000 6000 4000 0 500 1000 Time [days] Number of actions war I war II war III Number of actions per day on the time-line. The three coloured vertical stripes indicate war periods, the vertical line in-dicates the introduction of a major new game feature. Faction #1 Faction #2 Faction #3 Neutral 200 400 600 800 1000 1200 3 2.5 2 1.5 1 0.5 x 104 Time [days] Number of communication events (a) war I war II war III Number of weekly actions for different factions of players. Activity of players in-volved in wars is increased. (a) commu-nication C.
  • 10.
    CONCLUSIONS ² Intereventtime distributions in the Pardus universe are highly non-trivial. ² Presence of periodic patterns on different scales. ² Bursty dynamics. ² Different kinds of actions are characterized by different decay constants. ² Periods of increased activity in the Pardus due to history specific events. ² Physica A 419, 681-690 (2014) [arXiv:1407.2006]