Kasimir

"How about a game of pazaak, friend?" Illvin asked, casually, shuffling his deck. He wasn't sure he knew this interloper, but he figured he could do with playing a game of pazaak during his break. And everyone liked pazaak, didn't they? Not like those godless souls going on about poker or blackjack, or Lady Fortune forbid, some newfangled invention he'd never the least heard of known as sabacc.

I can stay in. Can't be extremely active as my everything has to recover from the exam-imposed deadlock but I can promise I will appear and vote every cycle Edited to add: And Szeth - no need to worry about me, it's good/mentally healthy for me to have one or two SE games where I'm not actually going to be going all out. If I wasn't certain I could at least commit to a low level of activity, I wouldn't stay in.

On brand But hey, we're getting there, I think. 40.9% isn't too awful, though there's the whole 'broken games' and 'mech' confounding factors, and so on, but that's another story. Granted, 'too awful' depends an awful lot on context, and compared to the stability of the early SE rates, it's not great. This is more or less my take. It's fine as an experiment, and I think it's absolutely worth giving a shot. I question the assumptions player inactivity is based on - I don't think Archer or I really have good data for this, and I am dubious that either of us are, just by intuition, correctly modelling the mindset of frequent inactives, which is where I think the dead doc discussion Archer and I had about actually collecting more data comes in - it's at least helpful if we get a better sense (qualitative? Invite survey/focus group?) of what drives inactivity and how to handle it. Basically we're both not frequent inactives so I have some doubt we model them quite right. By and large, I agree with Illwei the Nights serve a purpose, and I don't think it should be an either/or in terms of lively Day discussion or Night discussion, but will accept that most players want to take it easy at Night. My claim would be the point is you are supposed to discuss at Night, the rejoinder would be that players often don't/do not feel lynch pressure, and I think it more or less is a wash and at the mercy of discussion drivers. I think there's concerns as well about the impact on the Elim team, as Araris points out. Just less about bussing - ideally if the Village isn't driving at Night, then the Nights give the Elims a chance to get a breather and take stock, instead of going back to fending off the next lynch. Remember, if you're counting on a buoyed Village, you are also counting on an Elim team that has just suffered their first lost now having to defend each cycle instead of having the Nights in which to operate with a little less pressure. That feels like a two-shot in terms of their mentality, and puts them very much under siege. (My antipathy for Elim might be showing here.) I do think this feeds into a bigger issue where we just don't understand the causes of inactivity very well, much less how to tackle them. And this despite years and years of SE data on the matter.

GM flex, but thanks for running this and the bemused World Cup banter

Thank you Araris, thank you Mbappe

That's something that has to be checked, IMO. We did have inactivity in 2014 - made worse with fewer pinch-hitters and the fact games didn't always have filters. There's also the question of inactive new player Elims. There's a reason inactive Elim play used to be more rewarding too, strictly in terms of incentives. Arguably, though this is closer to 2015, AGs have been lost on inactivity, even. Not saying this is wrong out of hand but that I think it's worth doing a deeper dive into those two years as a comparator anyway! Sort of just see what the data says.

In LG85 you submitted a request to be Evil and didn't want to be Village back when I was doing the alignment experiment ¯\_(ツ)_/¯ Good work to Araris and Aman, both of you really hardcarried for your respective teams, IMO. And thanks for the PMs and the paranoia Araris >> Enjoyed paranoiding on you, JNV, and Alv!

Posted on the SE Discord in response to a question from @Shqueeves and a response from @Fifth Scholar but figured it was worth sharing in meta. The disclaimer: I did this sleep-deprived, grouchy, and fighting a cold while waiting for my DL epochs to finish running. So this isn't as clean/good as I'd like it to be. Shqueeves asked about current Village win rates, but it's really difficult to offer an assessment of Village win rates without - as always - putting the data into context. So here we go. Methodology This dataset was culled from the SE dataset here. This is a crowd-sourced dataset maintained by the community, though I believe earlier years were primarily captured by @little wilson and @Elbereth, with the original ur-dataset having come from an off-the-cuff project by @Wyrmhero somewhere between the LG5/MR1/LG6 boundary. Now, clearly to address Shqueeve's question, we need several things: A sense of how many games are run each year. If a year has fewer games, then even if the Village wins all the games that year (yeah, no, and that indicates something has gone very badly wrong!), we'll get the wrong impression that the Village is doing more poorly. This is not a good confusion to have. How many relevant (e.g. V/E, CON) games there are a year. Simply put, FFA and Faction* games are unlikely to have Village or Evil victories. As such, we should discard such data points as irrelevant to our question. % of Village wins: As suggested by the foregoing, having a % of Village wins makes a bit more sense. Raw numbers aren't helpful: we want them to be contextualised as a fraction/proportion of all relevant games. Making it a % makes it easier to display the results graphically. % of Evil wins: Sometimes, a game ends in a draw, or with a third party win. It is good to therefore also track % of Evil wins as a comparator, just for the sake of completeness. While the number of relevant/eligible games fluctuates each year, it is well within range. As such, a % is not so likely to be wildly misleading. Then, the SE dataset doesn't count years. As a result, I stipulated that a SE year is defined (with the reasonable exception of LG1) as the duration from the AG (or LG1) to the penultimate game of the year, i.e. the one right before the AG. Although the AG traditionally has sign-ups go up in late December, the bulk of AGs are typically played in January. As a result, I feel alright about taking the AG-pre-AG gap to be a SE year by definition. Challenges There are several problems when attempting to enrich this dataset: Inconsistency: While Wilson and El are probably the most consistent with applying the criteria for classifying a game as V/E, FAC, CON, EE, etcetera, it becomes very tricky when the criteria is being used by separate community members filling in the game. Simply put, it's important to note this is typical of a dataset that is being populated by multiple analysts at once. Having played some of the games in question personally, I have noticed V/E games misclassified as FAC games. In those cases, I have rectified the classification schema in my own enriched dataset. However: Size: The most number of games ever held in an SE year is 33. More often than not, there are at least 20 games. This isn't an especially small number, but it's not big either. This means inconsistencies in classification, i.e. dirty data, can really mess with the conclusions you want to draw. Missing Data: The years between AG5 to AG7 were particularly egregious in this regard, but AG4 also suffered from a number of incomplete classifications. As I was not very active in SE in this period and have been short on time, I brute-force imputed missing values in those years by assuming V or E wins indicated a game that is relevant to our interests. However, it is very possible that such games might (for structural reasons) be properly disregarded. 2019 was especially egregious with nearly 50% of classifications missing! Inactivity: As several SE players, including myself, have alluded to, SE has had periods of high inactivity alluded to as inactivity blights. During the Blights, some games have been declared a win or loss purely on the basis of the players who remained active. It is possible that inactivity blights might significantly skew the data within that period. Data Visualisations aka the actual stuff you're here for. Note that the 2022 data excludes QF64. I don't regard this to be a significant challenge as it is unlikely to significantly alter the overall %s for 2022. Percentage of Village Wins: In general, keeping in mind the caveat about 2017 to 2019 data, we can see that there has been minor fluctuations between 2014 and 2016. 2020 represents a very significant drop in Village win %s, which may as @Ashbringer and @Ooklil' the Wei suggest, be attributable to the COVID-driven influx of new SE players. If I have time subsequently, I can do a turnover visualisation on win proportions, but I have an exam tomorrow and this is a bit too much data preparation work for me now. I'll bring this up again later. This matches with anecdotal perceptions: between 2020 to 2021, @Araris Valerian commented on the need for a mechanic that might help the Village out by offering them a single Elim flip. This was later implemented as the Dragon's Fang mechanic of MR57. By 2022, the perception was that the Village was sufficiently holding its own, and Elim-side advantages may be required. It is worth noting that while we are currently on an upwards trend, the Village win % at the moment is 40.9% which is still not that great. The perception of the Village now being overly advantaged may be slightly inaccurate: in context, the Village is still underperforming. Percentage of Elim Wins: The perception that the Village is underperforming is confirmed here. In 2022, the Elims still outperform the Village. It is possible that this could be due to the prevalence of broken games (another factor to look at eventually) but by and large, it is clear there doesn't seem to be an especial 'Village advantage', but rather, Elim advantage is shrinking over time. Suggestions that games should be Village-skewed instead therefore seem premature. Interestingly, Elim win %s remain generally stable between 2014 to 2017, with a somewhat dip in 2018, which is dwarfed on this chart by the immense jump between 2019 and 2020. The picture is beginning to stabilise by this point and we are arguably seeing a return to normal. Comparing Elim Win % to Village Win %: Part of the point is to do a comparative picture, and on a comparative, we can see something very interesting: between 2014-2016, Elim win % and Village win % were exactly equal. That is to say, the Village and the Elims performed equally well. I think we can consider that an overall positive: ideally, we want there to be such a community state as the Village and the Elims performing equally well, or having equal win %s. This suggests that there is, on the whole, some form of meta-balance state prevailing. From 2017 onwards, we see divergence, with Elim performance remaining stable in 2017, but the Village vastly underperforming. This might be due to some corrupted data in that year, or the Village losing ground to third parties and other factions. Worth looking at, one way or another. Overall semi-parity is achieved in 2018 (Village skew) and 2019, but the situation is not great either. We have a significant divergence in 2020, with 2021 and 2022 representing slow/eventual convergence again, potentially as the community having absorbed new players begins to stabilise and develop a level of skill again. That being said, I don't think it is possible to make a strong comment here: recall that nearly 50% of 2019 data was unlabelled! I do expect that even with that uncertainty resolved, we will see Elims outperforming the Village in 2020, but it might not be as drastic. Avenues for Further Exploration: As indicated, we could try to look at the % of balanced V/E or CON games won. The difficulty with that is that we need a common enriched understanding of brokenness and overall balance. It is probably less problematic than I'm suggesting, but it will be an asterisk on any such analysis performed. We could superimpose the Elim/Village win % charts on new player influx per year. I do suspect we will see a spike in 2020 to match the Elim win spike, but it's good to actually refer back to the data to validate hypotheses. We could do something about player turnover: beyond new player influx, we have older players going inactive (e.g. on mission.) A turnover rate might better represent the changing community demographics and the impact on win rates. Inactivity can be compared to win rates, to see if there is any overall effect, particularly keeping in mind the inactivity blights. This would need further data enrichment. A problem with the theory that inexperience was a major factor in the 75-25 skew is that 2014 is a case study of a period in SE where by definition, everyone was a new player! A naive response or expectation is that an influx of new players distributes evenly: they may struggle with Village play but should also struggle with Evil play and thus be more catchable. While the opposite theory might be that it is easier for a new player to be Evil than Village, it's worth comparing 2014 and 2020 more closely to see what's going on, because 2014 defies this theory. Enjoy data

Odd to do one slightly shy of QF64, but I've been holding this one back until LG91 is over. A brief SE retrospective from my POV for 2022: good memories for me, and one moment of >:( Just thought it might be fun to make it a meme/stocktake thing.

Never go up against @Alvron's team when RNGesus is on the line? Brb re-reading with thematic music. Edited to add: Me this game when Alv, Araris, and JNV all claimed by N2: (No actually mass GM PM paranoia/breakdown resulted because it is clearly a lie, no way V!Chana connected me to V!Chana, V!JNV, and the Shaman at once...by N1...) Clutch shooting on multiple levels, Araris. I am filling in in my head an epic duel between Araris and Aman just because I can.

It's not the comment you think it is, but it's also not a comment I'm at liberty to elucidate on at this juncture.

:eyes:

Hmm. In the spirit of recent discussions, I am going to need to consider if I can remain in this game, and will try to offer some clarity by my Friday. While I submit the exam on Friday, I do have some backlogged reports that will probably burn up my weekend at least, and I dislike signing up for a QF where I can't commit to a basic minimum of activity that I can feel good about. Leave me on the roster for now, but this is a FYI that I've got to think and come back. Sorry Szeth, this looks like a fun game so I'll see if I can juggle stuff about :/

Very true Except you troll and part of the Elim wincon is fulfilled when you die, oy vey. But who cares about being responsible, the trolling is a once in a lifetime opportunity!

Thanks! Ngl this game looks like it's gonna be hard to be Inner Council. Who can resist trolling when you have a captive Elim audience?

Yeah, I don't really care about that. The point of my question was just to know if you'll be requiring email access. If I don't have to create a throwaway just to play this, I'm much more likely to stay on the playerlist

You can sign in to do it, right? I believe you'd still appear as an anonymous animal to others on the doc. I'm just worried that you don't appear anonymous in edit history or permissions history. But I also really don't see how you can selectively allow edit access to some and view access to others without requiring emails, so it feels like something that will need to rely on player integrity. RIP bro.

Eh, you basically have to do it inked anyway - you need the lubrication. Theoretically can wet the buffing stick/mylar as well but I think most videos like Pen Habit's generally recommend using ink instead of dipping it in water at each pass. Loupe feels necessary in case alignment shifts while you work on it.

I would definitely do that with a Wing Sung or a pen with a broken tine anyway. It's already broken or absurdly cheap, what's the harm. But I assume that needs some mylar as well to smoothen the edges after that?

Theoretically yes ._. But I actually have one more exam. It's just that it's 2nd Jan so lol soddit I guess? >>

How do you plan to swing the edit v. view access issue - player integrity? Feels somewhat revealing if players have to give you their Gmails, as this would immediately give a pool for who the spies and inner circle/winzik could be by checking the permission settings.

Signing up as Illvin Karrde, one of Starsight's many, many (human) docking crew who plays a mean game of sabacc poker in his off hours. A good way to celebrate the end of my exams >>

No, I was implying that if neither of us are aware of any LG coming up, it probably implies the answer is doubly, "No, the AG is next."

I mean, even if it were, Wilson and El have priority, followed by me...

Judging from the past two years, this is unlikely. And the current LG is also still running. Factor in the one week of sign-up time and the potential length of a standard LG and I don't really see a way to squeeze it in unless the mods are planning on pushing the AG back.