Originally published at: OCTGN Matchup Analysis, Part 1 - StimHack
Discuss the latest StimHack article by @Ajar here.
Originally published at: OCTGN Matchup Analysis, Part 1 - StimHack
Discuss the latest StimHack article by @Ajar here.
Very well done! It’s definitely a topic that merits extensive discussion, and I’m impressed with how much data you’re able to express within those easily discernible graphs.
I’m interested in how the matchup has improved against Weyland as well. I think Curtain Wall is death for Atman-oriented builds, and Hive forces Parasite installs. I’m also intrigued by the low winrate against NBN. Granted, NBN is an easy matchup for nobody, but I’d have thought Shapers would be a bit closer to 50/50. My guess would be there are good Shaper builds out there that are a bit too big-rig style and can’t properly address NBN’s fast advance (or that there is a large subset of Shapers that rely on resources, but I find that less likely).
Thank you for an excellent analysis, it was a great read.
I think that Kate’s advantage versus Jin was mostly thanks to C&C’s Clone Chip and SMC. The two of them allow Shaper to retrieve Deus X and continuously recurse it, stopping most of Jin’s net damage-related shenanigans.
May I ask whether you plan to share the code and the data? Maybe on github?
Since you’re using R and plan to re-run this analysis as more games are played, you might want to look into knitr, a reproducible research package. It might make your life easier.
This might just be me being a bad stats-person, but I’m wondering why you left out runners of less than 45 cards. While going by your numbers there’s unlikely to be a statistically significant number playing Chaos Theory (if even TWIWY and GRNDL are too low…) it still seems odd to limit the one side so differently. Is there a reason for it that I’m just missing?
Either way, this sort of analysis is excellent, and begins to answer some questions I’ve been wondering about for quite a while. I look forward to seeing the rest of the series.
In the vein of Kate vs Jinteki: Shaper just sucked in general before C&C. Clone chip, Deus Ex, and SMC are good, for sure, but even without playing those, you gained Pro Contacts and the entire Atman archetype.
Kate’s winrate actually increased pretty sharply across the board on the release of C&C. What’s more interesting to me is that her winrate against Jinteki continued to climb for most of the duration of the cycle, (despite not gaining very much). It’s possible that with Grim and Power Shutdown jinteki began to focus more on program trashing, which isn’t very good against Shaper in general, but that’s pretty heavy speculation.
It’ll be interesting to see how the loss rate for PE compares to the rest of the field over the spin cycle. ST gave us Profiteering and Clone Retirement and a wave of Jinteki decks that played very little ice and gained lots of bad publicity. Not a very good matchup vs Shaper, but I wonder if those decks just lost 70% of the time to the field in general.
@IirionClaus There’s a link to GitHub in the first section of the article, and I wrote the article in R markdown using knitr to update it whenever I changed something. I didn’t get quite so far as to use RWordPress to upload the article, but I did as much of the work as possible in a readily reproducible way. Here’s the GitHub link: GitHub - doorisajar/netreader: Processing and analysis of Android: Netrunner OCTGN game data.
@GreedyGuts Once I found that there weren’t enough games with CT to do her matchups, I filtered Runners on 45 as a convenient way of pruning illegal decks – it was just simpler than checking each ID’s deck size. I also didn’t do robust Corp deck checks for stuff like agenda count, since I saw from the work of people who did so that there wasn’t any noticeable effect on the results. As for the Corp deck size, TWIY* was right on the borderline of having enough data, so up until several days ago I was planning to include it in the analysis. That’s why the Corp deck size filters on 40. Between Harmony Medtech and Silhouette, I think it will be worth implementing more robust deck checks before the next data dump.
@Lysander Yeah, I’m right there with you on Weyland. It’s a very strange trend.
@mediohxcore I’m also with you on the PE matchup. I found it interesting that rather than a sharp increase with C&C, Kate’s winrate increased slowly, peaked, and then started declining as Jinteki improved over the last few packs. I’ve played Kate vs PE a fair bit and have had a lot of success, but the tools I win with are the tools that either predate C&C or came out in C&C.
Weyland gets worse when you learn how to play the matchup. Even when they were considered one of the top corps there was always the notion that they were “pubstompers” that get a lot worse against stronger players. It seems people have caught on to weyland’s tricks
Curtain Wall seems like a bit of a nonstarter to me-- it’s a substantially more expensive, often inferior Hadrian’s. Is anyone having trouble with this? Who is even playing it? I could perhaps see it in the 6x 5/3, 3x Hostile Takeover Ash/Troubleshooter/Oversight AI build, but that deck seems to have come and gone.
it just gets femmed anyway. curtain wall has the same issue of hadrian’s and wotan, paying that much money needs to do more than etr (unless you’re CI)
I think the Weyland numbers reflect the situation of Weyland in general, they where very strong going into Spin but it was the other corps getting the good stuff.
Also I dont really agree with the reasoning behind Atman and EtF, HB is probably the faction that has the best options to build variable ICE strength to counter Atman. The problem with HB builds have never been their Shaper match-up.
Excellent work btw!
sharpshooter was pretty bad for weyland.
“It’s also interesting that Weyland: Building a Better World went from being Kate’s second worst matchup before Creation and Control to being her second best matchup by Fear and Loathing. This is a shift I don’t have an obvious explanation for”
Clone Chip. Many people look at Weyland as “SE and nothing else” but their big strength is rig destruction as well. Clone Chip lets you rebuild the rig much faster and also enables playing Faerie in Shaper, which made importing Faerie much more viable.
Great writeup Ajar! Can’t wait to read the following parts.
One clarification needed, I’ll ask the question using an example on a picture since it’s easier for me to describe my doubts.
Just quickly before I head to work – the bar above a given pack reflects games played when that pack was the newest one available. so the ST bar covers the time from when ST showed up on OCTGN until MT showed up on OCTGN.
Totally makes sense.
The doubt came from Second Thoughts impact on Kate vs Jinteki PE matchup. Kate received literally nothing to work with in this datapack (apart from experimental and not really competitive voicepad decks) whereas Jinteki gained many good cards (clone retirement, profiteering, swordsman, restructure). One would think Kate would lose a bit more than in Opening Moves era…
Cool, completely missed the Github link. Thanks again.
W.r.t Jin, having cool new stuff sounds to me like a potential route to losing games, at least initially. A high win rate comes from practice, not from shenanigens- you’d get one or two surprise wins from Swordsman nuking Gabe’s Crypsis, but afterward they’ll be prepared and now your Swordsman is mostly a wasted Ice. Since Kate did not change much since C&C people had more time to practice.
I am in agreement with @IirionClaus. Crim was so strong for so long and the vastness and power of Shaper was not yet really known due to lack of experience and knowledge of how to maximize new cards, ect. Shaper is so versatile and unless the corps gets to rush a bunch of early agendas, as soon as I get past turn 10-12(ish), I feel like with a good shaper deck, I am going to win the vast majority of my games.
I like the point about Clone Chip being huge for the BaBW matchup. You’re absolutely right that Archer was a key cog in the BaBW decks that were prevalent around that time, and Clone Chip takes Archer from being backbreaking to a tempo/econ hit. Archer was still good, just not amazing, and that could easily be why Kate’s winrate vs BaBW climbed so steadily.
I also want to comment on the CT thing a bit more, since a few people on Reddit asked about it as well. I took a few minutes just now to change the Runner filtering to >= 40 cards and run all of the code again, just to triple check that excluding CT hadn’t affected anything. The player rating distribution was essentially unchanged (average 1429 vs 1428, stdev 148 vs 149). The distribution plot is virtually identical.
This implies that illegal decks are pretty uncommon – i.e. that not a lot of people are running 40-card Andy or Kate decks on OCTGN. That’s also useful information. (I later counted these games and found that in the 210k dataset there are 35 games played with 45-card IDs using decks that have 40-45 cards.)
Given all of that, I’ll simply use this new version of the dataset for Parts 2 and 3. It doesn’t change anything and illegal decks weren’t a problem anyway!
What follows are more details for the interested. Back when I initially looked at CT’s matchup numbers and concluded she didn’t have enough games to merit inclusion in the articles, I didn’t bother saving any numbers for her, so I played around with the data a little bit to look at them.
Including CT in the data resulted in the skilled players set going up to 695 from 678 – i.e. 2.4% of the skilled players were playing CT. But at least some of those players really like her, because the dataset of games played by the skilled players went form 71k to 81k games. I suspect (but haven’t confirmed) that the huge outlier player I noticed who has played 2k+ games is – or at least was – a CT fan.
CT looks pretty popular if you only look at total OCTGN games played:
RunID Games 1 Anarch | Noise 13610 2 Anarch | Reina Roja 2812 3 Anarch | Whizzard 6332 4 Criminal | Andromeda 12785 5 Criminal | Gabriel Santiago 14243 6 Shaper | Chaos Theory 9869 7 Shaper | Exile 1591 8 Shaper | Kate McCaffrey 14322 9 Shaper | Rielle "Kit" Peddler 4454 10 Shaper | The Professor 1148
However, CT’s play has dropped off tremendously since Opening Moves, which is why she doesn’t have enough data to merit doing the matchup trends:
RunID Pack Games 1 Shaper | Chaos Theory Trace Amount 4 2 Shaper | Chaos Theory Cyber Exodus 1350 3 Shaper | Chaos Theory A Study in Static 1827 4 Shaper | Chaos Theory Humanity's Shadow 597 5 Shaper | Chaos Theory Future Proof 1399 6 Shaper | Chaos Theory Creation and Control 562 7 Shaper | Chaos Theory Opening Moves 1835 8 Shaper | Chaos Theory Second Thoughts 430 9 Shaper | Chaos Theory Mala Tempora 304 10 Shaper | Chaos Theory True Colors 629 11 Shaper | Chaos Theory Fear and Loathing 751 12 Shaper | Chaos Theory Double Time 181
In contrast, here’s Kate:
RunID Pack Games 1 Shaper | Kate McCaffrey Trace Amount 425 2 Shaper | Kate McCaffrey Cyber Exodus 544 3 Shaper | Kate McCaffrey A Study in Static 1385 4 Shaper | Kate McCaffrey Humanity's Shadow 481 5 Shaper | Kate McCaffrey Future Proof 2036 6 Shaper | Kate McCaffrey Creation and Control 962 7 Shaper | Kate McCaffrey Opening Moves 3703 8 Shaper | Kate McCaffrey Second Thoughts 1152 9 Shaper | Kate McCaffrey Mala Tempora 606 10 Shaper | Kate McCaffrey True Colors 1261 11 Shaper | Kate McCaffrey Fear and Loathing 1513 12 Shaper | Kate McCaffrey Double Time 254
CT did pick up steam over MT and F&L, so it’s possible that by the time the next data dump comes out, she’ll have enough data to be included. Like TWIY*, she isn’t far off.
I actually think it would be quite interesting to see which players are loyal to their IDs and which players switch around a lot, and whether that’s correlated with player skill. I’d have to think about how to do that, though, and it certainly won’t be until after I get Parts 2 and 3 written!
Unfortunately, the OCTGN data doesn’t include deck lists. We won’t get that info unless db0 updates his Netrunner OCTGN plugin to save it. However, it would increase the amount of data stored per game by a factor of about 5, so I don’t think it’s too likely to happen.