[Planning] Mafiascum Site Championship Series

[schadd_]

sure i'll plan to mod some for this

[schadd_]

some thoughts

• it's probably not lost on you that one game/month is wildly optimistic. my 17p francisco lindor game opened for signups feb 3rd, filled mar 20, started mar 21, and finished may 19 (with 7 players surviving !!!). the micro from team mafia took a month and two weeks to finish. realistically you can take action to tighten these numbers as a mod: shortening deadlines, having backup moderation (and like setting gcal notifications and whatever lol) so that you get phases done closer to exactly when they're supposed to, et cetera. maybe the games fill faster, especially if there's this wad of money hanging over them, though i wouldn't count on it (my game also filled kind of phenomenally slowly in any case). i also think that signups for the next game should be open after the previous game goes down to 5 players or less, unless it seems reliable that signups for this happen faster than a week.
  
  somehow when i first read this i thought that the 5 best games counted instead of 7, which i actually think is preferable (in conjunction with keeping 12 games instead of 15). in particular, i think that this event works better if we aim to see it finishing sooner rather than laterㅡpeople leave the site and/or stop wanting to play, and the whole thing operates on the confidence of the players that "eh, i can stick with this thing for that long." the number 12 is conceptually glued to the idea that it would last a year (even if that literally could not happen ever on gods green earth) and i guess my instinct is that its just a better number. i imagine the thing really grinding to get done with 15 games.
  
  
• was about to type out something to the effect of "i think anonymous games are slow to fill but i haven't paid attention to any of them" and was aghast to find out that 13p anonymous mini themes seem to fill in about a week. still, i think it should be made easy for players to log
  n
  games played without having to play an alt game, because i imagine that some percentage of people just have no interest in them at all. there was a while where i didn't notice the alt game clause in the OP & i had a concern about the last few games, that people could openly complain about how people play when there is very cut and dry money on the line, which IMO necessitates like 2 or 3 or maybe 0 alt games, especially since perverse incentives in general will happen in every game regardless. i'm speculating blindly about how much of a chore they are for people so i would propose making a poll about this.
  
  
• i spent a lot of time thinking about point values. in short, my suggestions are: get rid of scoring for vig shot, all NKs other than 1 point for N1/N2, scum surviving at endgame; add 1 point for winning the game as either alignment; limit players to 10 games played out of 15 (or 7 or 8 out of 12), which can maybe be amended to "only the first
  n
  games count" if games struggle to fill. i'm just gonna click post now and plan to defend these later

[schadd_]

1) ...

sure

2)

mostly for the exercise (i dont do enough of this sort of thing lately...) i made an R simulation for this and got that there would be a tie between first and second place about 18% of the time with best 7 out of 15, vs. about 13.5% of the time with 5 out of 12.

Spoiler: methodology

• generate a game result: 50% chance to produce 0 + about 1.5 points, and 50% chance to produce 3 + about 3 points, capped at 9. i'd assume that a player who wins is likely to get more bonus points (for example i assume that "survives at endgame" doesn't include getting endgamed?)
  
• for 5/12: make 8 players who play 9 games, 5 players who play 8 games, 4 with 7, 3 with 6, and 2 with 5. i didn't think very hard about these numbers - i'm conceptually working off of 12 17-player games, and so this takes up about 80% of player-games and leaves 20% for people who just played a couple games for fun. no idea if it would work out like that, obviously there are large ramifications on the numbers based on how exactly it plays out
  
• for 7/15: 8 players who play 11 games, and then just 4 who play 10, 9, 8, 7.
  
• add the highest 5 or 7 scores from each player and check whether the highest and second highest are the same
  
• DO THIS TEN THOUSAND TIMES!!!!!!!!!!!!!!!!!!!

Code: Select all

```{r}
game_result <- function(){
  return (min(c(sample(c(0,3+rpois(1,1.5)), 1)+rpois(1,1.5), 9)))
}


guyvector <- function (x){ #player who has played x games
  ret = c()
  for (i in 1:x) {
    ret = append(ret, game_result())
  }
  return (ret)
}



ties1 = 0
ties2 = 0
for (p in 1:10000) {
  bmf = c()
  for (i in 1:8) {
    bmf=append(bmf, sum(tail(sort(guyvector(9)),5))) #generate 9 games; add up the 5 best results
  }
  for (i in 1:5) {
    bmf=append(bmf, sum(tail(sort(guyvector(8)),5))) #generate 8 games . . .
  }
  for (i in 1:4) {
    bmf=append(bmf, sum(tail(sort(guyvector(7)),5)))
  }
  for (i in 1:3) {
    bmf=append(bmf, sum(tail(sort(guyvector(6)),5)))
  }
  for (i in 1:2) {
    bmf=append(bmf, sum(tail(sort(guyvector(5)),5)))
  }
  if((sort(bmf)[21] == sort(bmf)[22])) {
    ties1 = ties1+1
  }
}
ties1

for (q in 1:10000) {
  bmf = c()
  for (i in 1:8) {
    bmf=append(bmf, sum(tail(sort(guyvector(11)),7))) #now we take the best 7 out of 11
  }
  for (i in 1:4) {
    bmf=append(bmf, sum(tail(sort(guyvector(10)),7)))
  }
  for (i in 1:4) {
    bmf=append(bmf, sum(tail(sort(guyvector(9)),7)))
  }
  for (i in 1:4) {
    bmf=append(bmf, sum(tail(sort(guyvector(8)),7)))
  }
  for (i in 1:4) {
    bmf=append(bmf, sum(tail(sort(guyvector(7)),7)))
  }
  if((sort(bmf)[23] == sort(bmf)[24])) {
    ties2 = ties2+1
  }

}

ties2


```

it's up to you what to do with those numbers or to reject them obviously, and also how much to value ties to begin with. there ended up being more ties than i thought and maybe that makes my proposal to knock away a bunch of the points-scoring things weaker. i guess ties aren't very scary to me; the thing that team mafia does for ties (?) and also that the big MU tournaments do exclusively to select the winners where the players determine an MVP sounds like at least a comparably strong way to recognize player skill as the other metrics. and of course it could be something the moderator optionally does if you are worried about the whole perverse incentive situation. or alternately, specific matchup record between the tied players when they were opposite alignments, or else something like finding the proportion of the points scored by each player's teammates vs their opponents (i.e. to figure out whether they had skillful teammates or opponents. i'd volunteer to do the tallying here)

sort of not convinced about the calculus of less weighty investment when there's more games but idk i'd say about that

3)

this time around i guess i'm more spooked by the idea of like "i should push x because they're the favorite to win. and

i

want to win." especially since it's not unheard of for people to seem to have a vendetta-based push on someone in the first place. i still think that's much less of a pernicious problem in game N-4 vs. game N-2, even if there's usually gonna be a frontrunner by that point, especially because the "best 7 games" thing makes it hard to compute how many points someone's gonna end up with. and i would choose to the line in the sand such that it's easier to opt out of playing an anon game while still getting the max games notched. sure that's easier with 15 games.

2)

My reaction to most of your suggestions about scoring is kind of a "but why?" so I'll wait to hear your rationale for it before responding.
[/quote]
i guess i try to think about these a lot and they're just a big black hole of consideration.

vig thing: naming vig as the only special power role worth awarding points strikes me as problematic. first of all there's the obvious thing where you get bonus points compared to the field by rolling vig. secondly, shooting mafia isn't the precise measure of whether someone uses vig skillfully/productively, many good players will also prioritize shooting non-PRs, people that were gonna get executed anyway (and this conceptually includes obvious mafia and guiltied mafia, who just end up being free exclusive points for the vig), and also deep wolves when applicable (which impacts the game hugely but isn't valued higher than any old mafia). and then also there's just like plenty of power roles that can be used skillfully or unskillfully.

nightkill: a large portion of nightkills in later nights are PRs, confirmed towns, etc. rather than good players. night 1 and night 2 it's often more like "i dont wanna play against this person" but is still sometimes because somebody claimed PR (which in turn happened because they got wagoned!)

scum survival: this discourages bussing of course, and also puts scum in the very harshly awkward position of having to argue with their teammates when there's disagreement there. granted, i can't remember almost any game i've seen where the wolves specifically chose to bus someone rather than doing it in response to their teammate getting wagoned, and they also almost never talk in great detail about it when they do. i don't watch team mafia games and maybe it's different there though. anyway, it can be good to bus a teammate and it doesn't always reflect poorly on a scum player if they died At Some Point, which is why i would remove the final survival point but begrudgingly keep the day-to-day survival (because it's rarely good for someone to go down very early in any case)

there's more to say but i'm gonna click post again

[schadd_]

there's also the problems with the fact that you can cheese out the points for voting correctly just by always making sure to hammer a wagon, and also there's a sort of buzzer race that happens if someone announces a guilty (in which case the first

n

players to jump in and vote get a point). vaguely i would think about penalizing people for voting town in some way (+2 points for voting scum and -1 point for every two town you vote out?) and then also like having a subjective way to discern when an execution is automatic