In Part 1, I covered my methodology in filtering and cleaning up the latest Android: Netrunner OCTGN dataset for matchup analysis, along with reviewing the matchups for Kate.
Before I dive into the rest of the Runner data, I want to address a question that came up about Part 1. When describing how I pruned the OCTGN dataset, I mentioned that I filtered Runner decks to 45 cards or more, which meant that my Glicko ratings were calculated almost entirely without Chaos Theory games. I did this as a shortcut method of pruning illegal decks, since I had previously checked the games played in recent data packs by each ID and verified that there weren’t enough Chaos Theory games in the Spin Cycle to give her the matchup treatment. However, I didn’t mention that in the article, which I should have done.
Some people expressed concern about Chaos Theory games being pruned before the Glicko calculations. I was confident in my work, but since I believe in open analysis, I wanted to make sure I addressed it in a clear way. Accordingly, I changed the deck size filtering for Runners to 40 or more cards, verified that this only added 35 games with illegal decks, and ran the code again.
The resulting Glicko distribution was virtually identical: previously, the average rating and standard deviation were 1428 and 149, respectively; with CT added back in, these changed to 1429 and 148.
In the skilled players cut – players rated more than one standard deviation above average – an additional 17 players were added. This means that only 2.4% of the top players were playing exclusively Chaos Theory. I did notice that CT players overall had played 9,869 games with her, so some players really like CT. However, the vast majority of those games were played prior to Second Thoughts, and as I originally observed, there haven’t been enough CT games per pack since then to show meaningful trends.
On a related note, another suggestion I received was to include some discussion of overall OCTGN Netrunner play trends so that the “games played” numbers have more context. It’s particularly helpful since I’ve noticed some large swings in overall OCTGN play during different periods, as these barplots show:
This OCTGN dataset was released shortly after Double Time became available, so I expect that there will be plenty of Double Time data in the next dataset. We can see that not many people played Netrunner in December 2013; apparently after Worlds we all took the rest of the year off. This means that the Mala Tempora data is somewhat suspect due to the low number of games played, as we’ll see when we look at the matchup data. These plots also show that the abnormally high game count during Opening Moves was due to the length of that window rather than a change in the number of games being played over time. Netrunner play on OCTGN is increasing this year, though, with monthly rates above anything seen in 2013.
Also, in Part 1 I noted that I wasn’t sure how to explain the increase in Kate’s winrate versus BaBW. An interesting comment from jacek_wieszacze in the discussion thread shed some light on it for me: Kate may not have gained anything that helped her against Scorched Earth, but Clone Chip really helped her against Archer (and even Power Grid Overload). BaBW frequently used timely Archer rezzes to open scoring windows, and Clone Chip shrunk those windows considerably.
Lastly, as a reminder before we get to the matchup plots, the winrate calculations are for games played by players with Glicko ratings more than one standard deviation above average. With all of that said, let’s look at the leading Criminals and Anarchs!
Ah, Criminals. Everyone’s favourite faction.
For much of the Genesis Cycle, Gabe’s winrate was so good against the field that the Haas-Bioroid: Engineering the Future trend is actually a bit hard to make out:
And the barplot version:
What’s most interesting to me about these trends is that Jinteki: Personal Evolution was Gabe’s worst matchup through most of the Genesis Cycle, although that isn’t saying too much since his winrates were still generally above 50% even versus PE.
The clearest trend visible on the line plot is the sudden drop in winrate versus NBN: Making News and HB: EtF when Opening Moves became available. Jackson Howard is the obvious likely culprit there, smoothing Corp draws and allowing Corp players to comfortably overdraw and discard agendas to get the cards they need. However, there was also another card in Opening Moves that Gabe didn’t really like at the time: Grim. Until the release of Knight in Mala Tempora, Gabe’s options for dealing with Grim weren’t ideal: Faerie, Parasite/Datasucker, Crypsis, Femme Fatale, or Mimic/Datasucker. None of those options efficiently neutralize Grim the way Knight does.
Knight in Gabe decks also works well against common “gear check” binary ice played in Corp decks that like to rush early agendas, and even more so when combined with Inside Job. In my view, that combination likely explains much of Gabe’s winrate rise versue Weyland: Building a Better World since the release of Mala Tempora. The other three core Corp IDs don’t generally lend themselves to the rush style, but it wouldn’t surprise me if Gabe’s winrate versus NBN: The World Is Yours were similar the BaBW trend.
Beyond that, there isn’t a lot to say: Gabe was generally outstanding in the Genesis Cycle and remains strong in Spin, with winrates hovering around 50% in three out of four core Corp matchups and almost 65% in the fourth.
That leaves Andromeda.
Andromeda was comparable to Gabe in the Genesis Cycle in terms of overall winrate – just slightly ahead against the field, as analyses by Alexfrog and Lysander have shown. But even in Genesis, there were differences between how our two Criminals performed against the core four Corp IDs. Gabe’s matchups trended together overall through much of Genesis, with only the BaBW matchup falling off toward the end of the cycle. Andy’s performance was less unstable versus EtF, MN, and BaBW, but her PE matchup wasn’t as good until Opening Moves. I’m skeptical that the 10% winrate increase came solely from John Masanori, but I expect he was a factor; drawing more cards against PE is hard to argue with.
It’s also interesting that Andromeda’s MN matchup was significantly worse than Gabe’s during Genesis, and didn’t show the same immediate Jackson Howard effect with Opening Moves as Gabe’s matchup trend.
Andy’s winrate versus BaBW – which started above 55% – has climbed fairly steadily over time. In contrast, while Kate’s winrate versus BaBW has also climbed since the release of Creation and Control, it went from roughly 40% at that time to ~55% during Fear and Loathing. Additionally, in the Fear and Loathing window, Kate was below a 50% winrate versus both EtF and MN, whereas MN was the only Criminal matchup of the four that was below 50%. However, Gabe was barely above 50% versus PE, with Andy at 55% and Kate at almost 60%.
There are a couple of interesting differences between Andy’s matchup trends and Gabe’s trends against the core four Corp IDs since the beginning of the Spin Cycle.
First, Andy’s MN winrate didn’t change with Opening Moves, while Gabe’s dropped about 15% and hasn’t recovered. Instead, Andy’s MN winrate declined later. With so few games played during Mala Tempora, it’s impossible to say with certainty whether her MN winrate dropped with Mala Tempora or True Colors, but the release of Sweeps Week in True Colors certainly didn’t help Andromeda.
Second, Andy’s winrate versus EtF hasn’t shown the same post-Opening Moves decline as Gabe’s. In fact, she won nearly 70% of her EtF games during the True Colors window before dropping back down to Earth in Fear and Loathing at the same time as Gabe’s winrate recovered to above 50% again. I don’t see much EtF in my local meta, so I don’t have a clear picture of why these trends have moved the way they have.
We all knew this tale of woe was coming.
A man who loves to destroy met his match in a man who loves to create. With credit to Leigh Alexander for coining the phrase, the toy-making genius of Jackson Howard has devastated our beloved hacker extraordinaire. The thin, frail lines we see plummeting after Opening Moves are a testament to how thoroughly Jackson Howard has crushed both Noise’s ability to win and his popularity among skilled players, although there are still enough games to look at most of the matchups.
Since the dark day when Jackson Howard was appointed to the Board of Directors of virtually every corporation, Noise has seen only one window in which he had a matchup winrate above 50% against any Corp ID, and that was in a matchup – PE during Fear and Loathing – with so few games played that a single strong player going on a run could be solely responsible for the change in winrate.
Noise has begun to show some signs of life against EtF, with his winrate climbing more than 10% from True Colors to Fear and Loathing, and his winrate is also climbing slowly against BaBW. Anarchs gained Keyhole in True Colors and Hemorrhage in Fear and Loathing; Hemorrhage is showing up in some tournament-winning Noise decklists from that time frame, although I doubt it’s solely responsible for Noise’s gains. But as with every other Runner so far, MN is one of Noise’s most problematic matchups as of Fear and Loathing. Fitting that the corporation that is home to Jackson Howard is the one wreaking the most havoc on Noise.
But could there be a saviour for the Anarchs? Like Kit, Exile, the Professor, and Chaos Theory, Whizzard hasn’t been played enough for the matchup trends to be consistently useful. That leaves us with the Red Queen. With only three packs of data, the line plot is less than helpful:
The barplots are slightly more useful:
While Reina has seen more overall success than Noise, her winrates still lag behind all of the non-Anarch IDs. Like Noise, her best matchup has been PE, but as we’ve seen, the other Runner IDs have also had success against PE. The relative lack of in-faction draw and tutoring ability hurts; in my experience, it’s common for Anarch decks to win or lose on the strength of the first 10-15 cards in the stack.
There are some promising cards in the Lunar Cycle for our Anarch friends; I certainly don’t think we’ve seen the last of them despite sub-50% winrates in virtually all matchups as of Fear and Loathing.
I’ve received a number of suggestions for other things to look at using this data, and have been maintaining a “to do” list since shortly before posting the first article. Once I’ve finished Part 3, I’ll be looking at this list and working on at least some of it until the next set of OCTGN data comes out. If you’re an R programmer and want to build on what I’ve done, my code is up on GitHub. If you just ideas for anything from new angles of approach to ways I can tweak the plots, post in the comments or message me – this is a conversation, and I’m always looking for new insights!
Continue the discussion at forum.stimhack.com