Recommend
10 
 Thumb up
 Hide
18 Posts

Android: Netrunner» Forums » General

Subject: Defining OCTGN Metas: ANR Communities Analysis rss

Your Tags: Add tags
Popular Tags: [View All]
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
I was intrigued by Kandiru’s OCTGN population work in a recent analysis thread; he was kind enough to share his formatted input data (from db0’s 100k game data dump), and I ran a related analysis of my own. It does appear that there are 2-3 distinct metas (using the simplistic definition of “people that play mostly with each other”) present in OCTGN play. The results are intriguing, but somewhat lacking for practical use. Any speculation about ways to meaningfully use this data is welcome and encouraged.

Here is a diagram illustrating the three primary ‘communities’ (network associations, based on matches played) comprising the core OCTGN player base, each represented by a different color:


Points (“nodes”) in this picture represent individual players, and lines between them represent matches. The lines are colored to indicate the community membership of the Corp player in a match. There are dozens of peripheral networks (between 2-200ish players) that I ignored for the purposes of this analysis; I also excised all illegal & suspicious game records (rough method here – I’ll refine those definitions in a later post with ‘real’ data analysis) for the cross-group comparisons.



The blue community (A) stands at 581 players and 28,783 games played. The red community (B) has 711 players and 24,994 games played. As would be expected in a worldwide online community, there’s a large degree of interaction between the two groups. I’ve marked the third, smaller (195 players, 5732 games) group as A2 since it appears to be more of a sub-group of A; there isn’t a lot of interaction between the A2 and B groups. Note that these communities are defined by the computer running specialized math based solely on matches played. Simply charting the networks defines them, but does not tell us why they are distinct. So let’s have a look at that.

Time zones (Americas / Europe) are the first possibility that comes to mind, and are about the only hypothesis we have a chance of testing given the limited personal information that OCTGN tracks. Using the Game Start Time attribute, I graphed the hour (0-23) that a game started in by community. Best information indicates that this time is recorded by a server in Greece, which is GMT + 2 or 3 hours. Here’s the graph (weighted for comparison between groups, so no y-axis):



The results are somewhat supportive of the time zone hypothesis. A and A2 definitely peak 5-8 hours later than B, which backs the time zone hypothesis, and would make B the European population and A the Americas. My best guess for A2 is that this is the tournament crowd and their circle of regular opponents; also possible is an East Coast-West Coast effect. Those two options also aren't mutually exclusive. I’m open to other interpretations.

But do they play differently, you ask? I ran the basic win/loss numbers and identities played for each group. More nuanced analysis (e.g. win/loss per ID) is probably time better spent elsewhere, since we’re just looking at the general population comparison here. Broadly speaking, the groups are very similar, but with a handful of interesting differences (which are minor, but the sample is large enough to achieve statistical significance; i.e. the differences can be ascribed to a factor other than chance).

Corp Play: Win Rate

A: 51%
A2: 52%
B: 48%

Runner Play: Win Rate

A: 52%
A2: 57%
B: 47%




Selected Highlights from Identities Played:

A (compared against A2):
Very similar Corp selection rates: A slightly prefers Personal Evolution, A2 slightly prefers Replicating Perfection
Very similar Runner selection rates: A prefers Chaos Theory, A2 prefers Kate

B (compared against A, A2):

HB: Less Core HB; more willing to give NEXT Design (“Guarding the Net”) a shot
NBN: Less Making News, much more TWIY
Anarch: Less Noise
Criminal: More Andromeda, at Gabriel’s expense
Shaper: More Kate & Kit




It was fun to play with data in a new way. I’ll post a more detailed analysis of the OCTGN 100k results once I have the time to address them. It will be slightly different from other similar ones in that it will focus on soundly defining groups of players for comparison (e.g. competitive vs casual) with defensible, common-sense criteria (spoiler: it ends up being more like OCTGN 60k results after cleanup).
13 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Steven Tu
South Africa
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Really interesting analysis, it hurts my brain looking at all those numbers... BUT.

Can I please get the beautiful diagram in high Rez? like 3000x3000 or something? It's very cool
3 
 Thumb up
0.25
 tip
 Hide
  • [+] Dice rolls
Konstantinos Thoukydidis
Luxembourg
Esch-sur-Alzette
Luxembourg
flag msg tools
badge
Avatar
mbmbmbmbmb
Quote:
The blue community (A) is largest, at 28,783 players. The red community (B) has 24,994 players. As would be expected in a worldwide online community, there’s a large degree of interaction between the two groups. I’ve marked the third, smaller (5732 players) group as A1 since it appears to be more of a sub-group of A;



I want to know how you've ended up with >20k players per side, when I've recorded just 7696 unique players o.O

The timezone is based on a central server's timezone. That server is in Greece afaik.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
James Solow
msg tools
mbmbmbmbmb
What sort of data analysis did you do?
Any sort of clustering and machine learning to get those three groups?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Drake Villareal
United States
California
flag msg tools
Avatar
mb
If I had to guess...

"B" group consists of newer players, people who tend to put things like "beginner" or "just learning" in the title, this draw others of a similar experience level to play, forming the weakest group, "B".

"A" appears to be average, regular players, the biggest group, and the group that connects the newbies to the pros.

"A1" appears to be the experienced players, the people more likely to put "advanced" in the title, OR simply wait to join matches of players they know to be good.

DbZer0 wrote:
Quote:
The blue community (A) is largest, at 28,783 players. The red community (B) has 24,994 players. As would be expected in a worldwide online community, there’s a large degree of interaction between the two groups. I’ve marked the third, smaller (5732 players) group as A1 since it appears to be more of a sub-group of A;



I want to know how you've ended up with >20k players per side, when I've recorded just 7696 unique players o.O

The timezone is based on a central server's timezone. That server is in Greece afaik.


7696 Unique players, but 59K unique matches? It adds up 29k+25k+6k = ~60K
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Konstantinos Thoukydidis
Luxembourg
Esch-sur-Alzette
Luxembourg
flag msg tools
badge
Avatar
mbmbmbmbmb
I doubt the connections are about experience, as online you don't really tend to play people of the same skill necessarily, unless you actively attempt it. Most people just play randoms.

If I had to guess, I'd say it's a timezone split. A being American timezone (largest), B being Europe + Africa and A1 being Asian + Australia.

EDIT: Fixed timezones
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Lysander
United States
Georgia
flag msg tools
Avatar
mb
DbZer0 wrote:

If I had to guess, I'd say it's a timezone split. A being US timezone (largest), B being Europe and A1 being Asian + Australia.


So you're going on record saying Netrunner is yet another game Asians best the world at?


....and let the great BGG debate begin!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
DbZer0 wrote:
Quote:
The blue community (A) is largest, at 28,783 players. The red community (B) has 24,994 players. As would be expected in a worldwide online community, there’s a large degree of interaction between the two groups. I’ve marked the third, smaller (5732 players) group as A1 since it appears to be more of a sub-group of A;



I want to know how you've ended up with >20k players per side, when I've recorded just 7696 unique players o.O


Sorry, that number is the count of game records between the identified players. I'll correct that once I can get to the computer with the source data again.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Evan
United States
California
flag msg tools
badge
Avatar
mbmbmbmbmb
Lysander1 wrote:
DbZer0 wrote:

If I had to guess, I'd say it's a timezone split. A being US timezone (largest), B being Europe and A1 being Asian + Australia.


So you're going on record saying Netrunner is yet another game Asians best the world at?


....and let the great BGG debate begin!



I'm trying not to be too literal about the spatial arrangement, but I'd have expected the Australasian meta to be a gradient offshoot of the European, rather than the American. Indeed, the time graph indicates that A1's daily cycle is more similar to B's than A's, but for some reason they're playing a lot more of their games against A?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Konstantinos Thoukydidis
Luxembourg
Esch-sur-Alzette
Luxembourg
flag msg tools
badge
Avatar
mbmbmbmbmb
From my experience, Australian timezones don't tend to sync good with EU. Not sure how much better it is with America.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
quarkyphysicist wrote:
What sort of data analysis did you do?
Any sort of clustering and machine learning to get those three groups?


No clustering or machine learning. Ran through Kandiru's data with Gephi, which implements the Louvain method for community structure. I set a few parameters and filters, but the software really does the heavy lifting computationally.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
Tuism wrote:
Can I please get the beautiful diagram in high Rez? :D like 3000x3000 or something? It's very cool :D


I'll see what I can do on Monday, will post it then!
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
mplain
Russia
flag msg tools
mbmb
Offtopic: I see an emergence of Data Leak Reversal decks, it would be interesting to know how often Corps lose by decking.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
mplain wrote:
Offtopic: I see an emergence of Data Leak Reversal decks, it would be interesting to know how often Corps lose by decking.


You can get an idea of that from the Win Rate charts, just look for the row that says "Deck Defeat". I'd guess that particular metric can be generalized to any player population.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Matt
Australia
Newcastle
NSW
flag msg tools
Avatar
mbmbmbmbmb
DbZer0 wrote:
From my experience, Australian timezones don't tend to sync good with EU. Not sure how much better it is with America.


Anecdotally, I generally end up playing against Central/West Coast US players if I'm playing outside AU/NZ (I'm east coast AU). Most of my EU interaction comes over the weekends.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
Tuism wrote:
Can I please get the beautiful diagram in high Rez? :D like 3000x3000 or something? It's very cool :D


Will you settle for 2796x2530?

I'm having problems with the image uploader, but send me a PM and I can get you the file.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Steven Tu
South Africa
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Miaowara_Tomokato wrote:
Tuism wrote:
Can I please get the beautiful diagram in high Rez? like 3000x3000 or something? It's very cool


Will you settle for 2796x2530?

I'm having problems with the image uploader, but send me a PM and I can get you the file.


Got your PM, thanks! Had a super busy weekend so I haven't been watching the forums
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
Worked out uploader issues. Here's a link for the "smaller" (1864 x 1687) version for anyone else interested.

1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.