Recommend
23 
 Thumb up
 Hide
19 Posts

Android: Netrunner» Forums » General

Subject: When the Best Play the Best: Examining High-Elo Matchups rss

Your Tags: Add tags
Popular Tags: [View All]
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
Earlier this week we looked at some faction performance data over all skill levels. One major question that went unaddressed there was "What happens to these results at high skill levels?"

Before jumping to the charts, here's what the general OCTGN population Elo distribution looks like:



Runner ratings and Corp ratings were calculated separately (details), which is why you'll see different totals. For this analysis, I used only matches where both players were above a 1650 rating. This reduces the dataset to 5k games, of the original 105k from the previous post.

In general terms, the highly-rated groups play very similarly to the general population. "1650+" games average 15.09 turns with a standard deviation of 5.8; "General Population" games average 14.87 turns with a 5.92 standard deviation. The highly rated group plays approximately 20% faster (Avg 23.4 min vs 29.1 min) Their games play out in very similar ways; here's how Gen Pop games end (Blue is Runner victory, Red is Corp):



and here's the 1650+ results:



1650+ runners score agendas, flatline early, and put their pants on one leg at a time, just like you. Except, once their pants are on, they beat the living bezeesus out of unsuspecting corporations.



(it's the internet, pants are not actually guaranteed, come to think of it)

I've included the # of games for each data point in these charts to reinforce that some of them are based on very few games; know that the Variance Daemons may have nudged them a bit from their theoretical "true" position.

Runner performance:



Corp performance:




'Nuff said, right?

Interactive version is also available. Switch between tabs to see all pages.
37 
 Thumb up
1.75
 tip
 Hide
  • [+] Dice rolls
Bryan Blumklotz
United States
Portland
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
Bryan,

Always informative. Hope to see you at the tables this weekend for our back to back Store Championships...
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sasha F
United States
Kansas
flag msg tools
Avatar
mbmb
Miaowara_Tomokato wrote:
In general terms, the highly-rated groups play very similarly to the general population. "1650+" games average 15.09 turns with a standard deviation of 5.8; "General Population" games average 14.87 turns with a 5.92 standard deviation. The highly rated group plays approximately 20% faster (Avg 23.4 min vs 29.1 min)


I can't quite tell if this tells me a lot or nothing at all. A standard deviation of 1/3rd the average is absurdly high. According to this, 68% of all games recorded on OCTGN take about 9 to 21 turns... clearly the turn length is highly dependent on the decks involved and it would be very hard to interpret any sort of win correlation to game duration (in turns). Basically from this I understand, "take as long as your deck needs and don't pay attention to what turn it is." That's pretty comforting to a newer player.

I'm not terribly surprised that better players play faster. I would be interested to see if play speed is more strongly correlated with experience (number of games played) or ability level (ELO). The question I'm trying to answer with that is whether better players just realize their optimal moves quicker (likely) or if playing faster in general helps you win (possibly throwing off the opponent's tempo). Learning is fun!
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Anton R.
Russia
flag msg tools
Avatar
mb
Heh look at Criminal drop after power shutdown was released.
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Christopher MacLeod
Canada
Moncton
New Brunswick
flag msg tools
mbmbmb
gozik wrote:
Heh look at Criminal drop after power shutdown was released.


Mwahaha

Playing a variant of your Power Shutdown Jinteki the other day, by Turn 20 the poor Gabe had all 3 Faeries, 2 Corroders, a Zu, a Mimic, and a Femme all in his heap through a combination of Power Shutdowns, Grim, and random net damage.

Then I killed his Crypsis with Swordsman
6 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
Squash957 wrote:
I would be interested to see if play speed is more strongly correlated with experience (number of games played) or ability level (ELO). The question I'm trying to answer with that is whether better players just realize their optimal moves quicker (likely) or if playing faster in general helps you win (possibly throwing off the opponent's tempo).


Runner Correlations

# Games <-> Rating: 0.45 (strong positive - they increase together)
Rating <-> Avg Duration: -0.14 (negligible negative)
Avg Duration <-> # Games: -0.28 (moderate negative - duration decreases as games played increases)


Corp Correlations

# Games <-> Rating: 0.38 (moderate positive - they increase together)
Rating <-> Avg Duration: -0.27 (moderate negative - duration decreases as rating increases)
Avg Duration <-> # Games: -0.26 (moderate negative - duration decreases as games played increases)


So... people tend to play faster the more games they've played, and higher-rated Corp players also tend to play faster. The latter effect isn't really there for runners; potentially because Corp decks will stick to a plan, and get better/faster at executing that plan on their turn, where runners need to constantly re-evaluate the board position to make their decisions.
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Peter O
United States
Oakland
California
flag msg tools
Avatar
mbmbmbmbmb
I want to know what the super-secret BWBI tech of the creation and control era is, and why no one shared it!
surprise
(And don't talk to me about sample size, I know a conspiracy when I see one)

 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Alex Rockwell
United States
Lynnwood
Washington
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
tranenturm wrote:
I want to know what the super-secret BWBI tech of the creation and control era is, and why no one shared it!
surprise
(And don't talk to me about sample size, I know a conspiracy when I see one)


It appears that you already know the answer.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Alex Rockwell
United States
Lynnwood
Washington
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
gozik wrote:
Heh look at Criminal drop after power shutdown was released.


Recently I played a game where I was Jinteki PE with Power Shutdown recursion. At the end of the game, my opponent had no deck left, and the entirety of his rig was a Garrote.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
spags
United States
Ill Mil
Wisconsin
flag msg tools
badge
Avatar
mbmbmbmbmb
Tip left before I even read it.

I know it will be that good.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Loren B
Uganda
msg tools
Avatar
mbmbmbmbmb
Alexfrog wrote:
gozik wrote:
Heh look at Criminal drop after power shutdown was released.


Recently I played a game where I was Jinteki PE with Power Shutdown recursion. At the end of the game, my opponent had no deck left, and the entirety of his rig was a Garrote.


I would love to see what that deck looks like...
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Anton R.
Russia
flag msg tools
Avatar
mb
Alexfrog wrote:
gozik wrote:
Heh look at Criminal drop after power shutdown was released.


Recently I played a game where I was Jinteki PE with Power Shutdown recursion. At the end of the game, my opponent had no deck left, and the entirety of his rig was a Garrote.


About a half of my games ends like this. But sometimes I have no deck neither and scores my last point at last possible turn I still never didn't decked myself but it's totally possible in this deck, especially vs shapers.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sasha F
United States
Kansas
flag msg tools
Avatar
mbmb
Miaowara_Tomokato wrote:
Squash957 wrote:
I would be interested to see if play speed is more strongly correlated with experience (number of games played) or ability level (ELO). The question I'm trying to answer with that is whether better players just realize their optimal moves quicker (likely) or if playing faster in general helps you win (possibly throwing off the opponent's tempo).


Runner Correlations

# Games <-> Rating: 0.45 (strong positive - they increase together)
Rating <-> Avg Duration: -0.14 (negligible negative)
Avg Duration <-> # Games: -0.28 (moderate negative - duration decreases as games played increases)


Corp Correlations

# Games <-> Rating: 0.38 (moderate positive - they increase together)
Rating <-> Avg Duration: -0.27 (moderate negative - duration decreases as rating increases)
Avg Duration <-> # Games: -0.26 (moderate negative - duration decreases as games played increases)


So... people tend to play faster the more games they've played, and higher-rated Corp players also tend to play faster. The latter effect isn't really there for runners; potentially because Corp decks will stick to a plan, and get better/faster at executing that plan on their turn, where runners need to constantly re-evaluate the board position to make their decisions.


Very cool, thanks! Want to get better at Netrunner? Just play more games!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Gregory Pettigrew
United States
Massachusetts
flag msg tools
mbmbmb
tranenturm wrote:
I want to know what the super-secret BWBI tech of the creation and control era is, and why no one shared it!
surprise
(And don't talk to me about sample size, I know a conspiracy when I see one)



I imagine it was me testing this deck.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
ed noid
msg tools
What K factor(s) did you use on your ELO calculation?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
ednoid wrote:
What K factor(s) did you use on your ELO calculation?


If ((PlayerRating < 2100) Or (OpponentRating < 2100)) Then
K_Value = 32
ElseIf ((PlayerRating < 2401) Or (OpponentRating < 2401)) Then
K_Value = 24
Else
K_Value = 16


So, effectively, 32. The conditions for lowering to 24 or 16 weren't ever triggered.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
ed noid
msg tools
Do you have a processing script/query for the ELO you'd be willing to release? I've found myself with some time on my hands and I'd like to run the OCTGN data through a few models, but my test runs in R are showing that the player id's are completely dominating my models. Basically uncorrected player skill counts for so much in everything I've run that nothing else shows up. Bucketing via elo would help I think.

I'd rather not have to work through a complicated ddpy/aggregate in R just to get the elo's in.

 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bryan Goodwin
United States
West Linn
Oregon
flag msg tools
badge
Avatar
mbmbmbmbmb
ednoid wrote:
Do you have a processing script/query for the ELO you'd be willing to release? I've found myself with some time on my hands and I'd like to run the OCTGN data through a few models, but my test runs in R are showing that the player id's are completely dominating my models. Basically uncorrected player skill counts for so much in everything I've run that nothing else shows up. Bucketing via elo would help I think.

I'd rather not have to work through a complicated ddpy/aggregate in R just to get the elo's in.



No problem; it's in Access VBA, though. Send me a PM with your email address and I'll send you what I have. Alternatively, one quick 'n' dirty workaround might be to just grab the calculated ratings and import them.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
ed noid
msg tools
sent pm. I already pulled in the ELO's for the high ranked players, but I'm interested in running a few analysis (ancova comes to mind) that control for skill, and in that case everyone's ELO would be helpful. Specifically I'd love to run an ancova on win rate vs faction and help put some of this data vs experience infighting to bed.
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.