Connections between Wormhole Regions in zKillboard Data

  1. zKillboard data from Jan 2008 to Dec 2019
  2. A person “jumps” when he/she appears in two kill mails that are within 10 mins apart.
  3. The probability to jump from A to B is (#jumps from A to B) / (#jumps from A to anything)
  4. Blurring makes the picture more clear. Image without blurring has lots of very tiny dots.

Wormhole region matters. For example,

  • B-R00004 (class 2) connects to class 2 and 3 regions.
  • B-R00005 (class 2) connects to class 2 and 4 regions.

Probabilities are low because:

  • Most capsuleers stay in the same wormhole for a long time. The figure hides the probability to stay in the same system.
  • When capsuleers die, they might re-spawn in known space. The figure only describes wormhole space.
  • A capsuleer might not kill two ships within 10 minutes.

Enjoy before CCP changes this :smile:
By the way, do you know where can I find more studies about exploration?

Nice graph! As expected, it’s mostly symmetrical. But not entirely.

I don’t live in wormholes myself, but I wonder if someone who is at home in wormholes has a possible explanation for the parts that aren’t symmetrical.

I am not sure at this point.

  • The figure is roughly symmetrical along the diagonal.
  • The bottom half represents a capsuleer generates a kill mail in a higher class wormhole system followed by another kill mail in a lower class wormhole system within 10 minutes.
  • The upper half represents a capsuleer generates a kill mail in a lower class wormhole system followed by another kill mail in a higher class wormhole system within 10 minutes.

Here is my guess.

  1. Suppose capsuleers who often win fights would live in higher class wormhole system.
  2. When they win a fight at their home wormhole system, they might jump to a lower class system and win another fight within 10 minutes. After that two things can happen: A. They move back to their home wormhole system, don’t find a good target, and log off. B. They move to an even lower class wormhole system and win another fight within 10 minutes.
  3. On the other hand, capsuleers who often lose fights would live in lower class wormhole systems. After they win a fight against a capsuleer or npc in a lower class system, they might go to a higher class system with a low probability. If they do that, they have a relatively high chance to get killed within 10 minutes. After that, they respawn in known space and would not immediately try this again.
  4. The combined effect is that the bottom half has brighter dots, which means higher probability.

Another possible reason is that capsuleers that win more fights have a higher chance to register at zKillboard and live in higher class wormhole systems.

The asymmetry is less pronounced when the jumps are between high class (classes 4, 5, and 6) wormhole systems. I guess that is because the capsuleers at both sides of the fight have roughly equal chances to die.

There are lots of things to check. But it is possible to build a model of cats versus mouses with the zKillboard data.

The Codes to the left are the names of the systems or types of systems?

How did you account for kills in same wormhole 10minutes later?

Did you write the logic to exclude those?

One would expect a diagonal line from top left to bottom right to represent that probability otherwise.

Also think of it less in terms of distribution and more in terms of horizontal/vertical.

The horizontal probabilities represent a specific classes’ total possible links.

A horizontal line represents a specific class was linked from.

The figure actually counts the behavior of capsuleers instead of the probability that a wormhole connects two system. To count the probability that a wormhole connects two system, I need to make a statistical or Bayesian model to guess

  1. what happens between kills in zkillboard and
  2. when do wormhole spawn and despawn.

1 means making histogram of time between kills in each system. Or trying to automatically identify different behaviors of capsuleers like wormhole rolling, exploration, pvpers, etc, and make statistical models. I think 2 is easier… Maybe I can try to use a Poisson distribution.

I guess there are correlations between connections between wormholes systems. Human behavior is not random. Capsuleers can roll wormholes that they don’t like.

While I set the diagonal to zero in the figure, I count the jump from a system to itself in the normalization. When a player makes 2 kills in system A and then 1 kill in system B all in the same 10 minutes, the program would record 1 jump from A to B and 1 jump from A to A. The program only memorizes the last occurrence of each player that are within 10 min from now.

I think:

  1. A wormhole always connects two system.
  2. A wormhole doesn’t favor a particular direction.
  3. Capsuleers might favor a particular direction.
  4. The figure describes behavior of capsuleers instead of the probability that a womhole connects two systems.
1 Like

Code to the left are regions of the wormhole systems.