Friday, 27 September 2013

Psychographics // 100,000+ OCTGN Games Analysed // Size Matters

I don't plan on making a living from milking the OCTGN data from the first 100,000 games of Android: Netrunner, but there is one last blogs-worth of data I can throw out which I think has some interesting findings.  Some of those findings reinforce the 'accepted wisdom' while others challenge the status quo.  

It's not a topic that gentlemen often talk about in public so gird yourselves - we're going to talk about size.


I already touched on the subject of deck size for Corporations in an earlier post and I'm going to return to it here with a wealth of OCTGN data to talk about.  Most Corp Identities have a minimum deck size of 45 cards, and for those identities there are three deck sizes that make sense as they play to different strengths and weaknesses:

45 cards - High Agenda density, Highest consistency
49 cards - Low Agenda density, Average consistency
54 cards - Lowest Agenda density, Lowest consistency

The tradeoff being made is being how many Agenda points you have per card in your deck, from a minimum of 0.41 (22 Agenda Points in 54 cards) to a maximum of 0.47 (21 Agenda Points in 45 cards), which might not sound like much but it's a 15% difference.  The opposite side to the change in Agenda density is that a smaller deck means you see your best non-Agenda cards more often (you'll see a card 22% more often in a 45 card deck than you will in 54 cards).

Smaller deck sizes suit a quick Agenda-scoring strategy, or an aggressive Tag punishment strategy that needs to see things like Scorched Earth.  Big deck sizes allow the Corp to get by without so much central server defense because the Agendas are spread more thinly in R&D.

Enough preamble, let's look at OCTGN and see what we know.

Corp Deck Size vs Win Rate


This data was pulled for Creation & Control games for all IDs except NBN: The World Is Yours, for all players.

First of all, you can ignore pretty much all the data points that aren't 45, 49, or 54 cards simple because they come from very little games - 94% of all Corp decks played one of those three deck sizes.  Next, you can see that the win rate was highest for 49 card decks (46%) while 54 cards (43%) and 45 cards (41%) lagged behind.

Where things get more interesting is when you cut the data down to just Corp players with win rates over 55% (so above-average Corp players).


Three things crop up here, all of which are revealing:

1) most of the players using 45 card decks are below-average, as the % of players halves in the >55% group. BUT the good players who are playing 45 cards seem to have the best records (nearly a 75% win rate!). The sample size is small (112 wins from 151 games) so it could be an anomaly... we'll come back to that in a moment.

2) The % of players using 49 cards increased to 89% in the >55% players. Most good players use 49 cards.

3) 54 card decks are not significantly disadvantaged from 49 card decks, which is interesting. The sample size is small but larger than it is for 45 cards (208 wins from 332 games).

The 45 Card Decks

I just said that the 45 card decks delivering the best win rate could be an anomaly, so let's have a quick look for that to see if a single set of results are throwing off that sample's overall performance.


Across the four most heavily played Corp Identities played at 45 cards the pattern is really consistent - they ALL performed better at 45 cards than they did at 49.  The performance at 54 cards was less uniform, with NBN: Making News seemingly strongest at 54 cards, and Haas-Bioroid: Engineering The Future distinctly weaker at 54 cards.

There is a huge caveat here: the sample size here is becoming very small.  From an initial sample of 100,000 games we've cut everything before August, everything where the Corp player didn't win more than 55% of his games during August/September, and we're focusing on a deck size that only 2% of players used - a grand total of only 151 games are left.  That makes these results unreliable, but the fact that it was repeated for all Corp IDs does lend some credibility to the theory that 45 card decks are strong at the moment (as I first posited a few weeks ago HERE).


Something For The Runners

Runners don't have to worry about Agenda density so deck size is much easier for them - the smaller the better.


Pretty obvious, but I'm happy to state the obvious now and then.


Conclusions

This hopefully ends my trawling through the OCTGN data dump and I can move on to new topics but I hope that you've all been able to learn something from this analysis, both the earlier posts about Corp and Runner identities, and this one about the tricky subject of size.

In summary I think the Corporation deck size question is at a very interesting place.  When players come into Netrunner they often bring preconceived notions that the best things come in small packages and stick to 45 cards.  The results from the above-average players shows that players migrate to 49 cards as they becomes more experienced and see other players running 49 cards.  What now seems to be happening is a third stage in the deck size evolution, with players finding ways to produce results at 45 and 54 cards that match or exceed the 49 cards.

So if you have a 49 card deck, will making it 45 or 54 cards make it better?  No, probably not.  You'll need to change what your deck is trying to do to get the most from that transition.

But at 45 or 54 card decks inherently better than 49 cards?  No, once again probably not.  It's possible that the players getting good results at 45/54 cards were getting above-average results at 49 cards as well.  What I think you CAN take from this is that 45 and 54 cards are now realistic options, and that makes for an interesting range of deckbuilding choices.