VCA Stats

[goodmorning]

I've been working on a project for a little while which is still not quite done, but I figure it's about time to share it.

I've got stats now for Large Normals, Opens, Micros, and Newbies, but since it's 3am and I'm high on exhaustion only Newbies are going up right now.

And yes, it's on google spreadsheets, ok, I don't want to talk about it.

Newbie VCA
I've only analysed the 69 Newbies that have ended this calendar year, pausing at the oldest non-finished Newbie game.
Teaser stats: 31% of lynches in these Newbie games have no Scum on them at all. About 3/4 of these are on Scum and 1/4 on Town.
The location of votes does not appear to be statistically significant.
EDIT Nov 9: 85 Newbies, now, but no significant changes in these teaser stats.
EDIT Jun 2: Approaching 150. Now down to 28.21% all-Town, with a 70-30 split as to who they're on. 13.61% of all D1 wagons in Newbies are all-Town on Town - around 1 in 7.

Have done no analysis on the rest and if/when I do it'll probably be quite limited due to the massive # of possible roles and large variation in wagon size/day.
Large Normal
Open
Micro
EDIT: Nov. 9: Have done some small amount of analysis for Large Normals and Opens, mainly in #s of all-Town wagons.

I'm still debating whether I should do Minis together or separate.
Large Themes will probably be last - they'll be more labor-intensive to input due to not being numbered.

[TierShift]

I want to see if second votes will continue to be more town-dominated.

Looks interesting to point out trends but there's nothing that catches the eye as of yet.

[EspeciallyTheLies]

So 69% of them do.. that's pretty significant imo and validates VCA as a method. But newbies are so different from other games. I've personally had the best results from 13p+ games.

[TierShift]

If my math is correct, 92% of town lynches have scum on them.

You can indeed work with that.

Edit: wrong, recalcing

[TierShift]

What's the percentage of mafia lynches?

[goodmorning]

What do you mean, percentage of Mafia lynches?

[goodmorning]

If my math is correct, 92% of town lynches have scum on them.

You can indeed work with that.

Edit: wrong, recalcing

Roughly, but that means around 2 in 25 don't. That's about twice per 8 games, if we assume the average game has 3-ish lynches.

EDIT: Math says the average game of data to this point is 3.41 Days.

[zoraster]

33% of town lynches have both mafia on the wagon.

[goodmorning]

Oooohh, I should do that stat too.

I just have to be careful not to overload it to the point where I won't update.

EDIT: Got that 20% of lynches in Newbie games had both Scum on, most of which were (unsurprisingly) on Town.

So with 20% having both on and a little over 30% having both off, 1 on 1 off happens a little less than 50% of the time.

[Cogito Ergo Sum]

So 69% of them do.. that's pretty significant imo and validates VCA as a method.

I'm pretty sure that random votes would get you higher percentages than 75% in pretty much all scenarios with at least 2 scum (you get lower here because of scenarios with just 1 scum, I imagine). It's literally not significant.

VCA has always just been the fallacious reasoning of taking a reasonable position such as "there is an average amount of scum on this wagon", combining that with information from the flips, then resolutely ignore the fact that this new information changes how reasonable your original position and pretend it's still likely to be true. E.g. if you're playing a newbie and everybody on a certain wagon but player A flipped town, then player A isn't 69% likely to be scum; the fact that everybody else flipped town is very strong evidence in favour of the fact that that wagon might've actually been all town.

[zoraster]

sorry, mine was Day 1. It would take more time to see how many lynches had both scum AND there were two scum left in the game.

[goodmorning]

sorry, mine was Day 1. It would take more time to see how many lynches had both scum AND there were two scum left in the game.

That would explain it.

[Lowercase]

If you do the calculations, 92% is only 3% higher than the expected percentage of town lynches with scum on them. Even if this were significant, the difference is far too small to be useful in scumhunting.

However, in the scum lynches, the first two slots were always town. Though the sample size in this case is only 16, this is still about 92% significant and common sense would imply that the trend is real. About 44% of these lynches had scum on them, which is close to what is expected if scum avoid the first two slots but have no preference otherwise. This probably isn't much use since after a correct day one lynch, town PR's tend to clean up anyway.

Do you have any stats for stuff like counter wagons, OMGUS, not voting, or self-voting?

[Lowercase]

33% of town lynches have both mafia on the wagon.

Unless I messed up, the expected value for random voting (assuming no self-votes) is that about 35% of town wagons have both scum on them. The take away here seems to be that wagon analysis on town lynches in newbies on D1 is completely uninformative if you only consider wagon participation.

[zoraster]

Yeah, I get 35% too. 5/8*4/7, right?

The statistic of 92% or whatever (haven't confirmed this number) with at least one scum on it is probably more significant.

[goodmorning]

@Lowercase: I don't have the time to do more than analyse votecounts, and counterwagons don't always happen when a mod is around to record them.
I considered stats on not voting but decided that the work outweighed the reward.
I've definitely thought about including self-votes and may go back and replace the alignment with the word "self," at least in Newbies, but the truth of the matter is that the majority of selfvotes on lynch wagons are selfhammers and the selfhammers are about 75-25 Scum.

[TierShift]

Guys, my 92% of town lynches have mafia on them is wrong. The actual percentage is lower.

I took the amount of town lynches with no mafia on them (1/4*31%~8%) and divided that by all town lynches, for which I took the lynches with scum on it+townlynches without scum (69+8~77%). That is wrong, though, because there are mafia lynches with mafia on them, which now artificially raises the result. So to correctly calculate this, I need the percentage of town lynches.

Even if I calculate it like this, it's 90 instead of 92%. If we know the amount of bussing, I can give the accurate percentage.

[Lowercase]

My calculations for at least 1 scum on wagon is (6*5*4 + 3*6*5*2)/(8*7*6) = .89. Slightly more significant, but even if it were exact it is too weak to draw a meaningful conclusion in-game.
p-edit: I actually counted myself, though I could be wrong. Looked like 50 town lynches with 6 no-mafia town lynches. Either way, it is incredibly close to random. Also, the random chance of lynching scum is about 22% compared to this sample which has about 24% scum lynches. I'm forced to conclude that everyone sucks at mafia and we should just be rolling dice and discussing whose number is better.
edit: Oh yeah, 6/50 = .12. So the real percentage for >1 mafia on wagon is 88%.

[TierShift]

I want someone to calculate town EV if all newbie replacements are PL'd. I think town winrate would shoot up by 5%.

So stop scumhunting, lynch replacements, thank you.

No throwing dice.

[Lowercase]

If the relative replacement frequencies for scum vs town hold for just day one (they probably don't), a PL on a newbie replacement has about a 28% chance to hit scum. I don't know whether it's worse to replace early or late, but it does seem you would probably be better off if you PL'd newb replaces.

There's also the added bonus that you are unlikely to lynch a cop, tracker or BP.

[goodmorning]

Guys, my 92% of town lynches have mafia on them is wrong. The actual percentage is lower.

I took the amount of town lynches with no mafia on them (1/4*31%~8%) and divided that by all town lynches, for which I took the lynches with scum on it+townlynches without scum (69+8~77%). That is wrong, though, because there are mafia lynches with mafia on them, which now artificially raises the result. So to correctly calculate this, I need the percentage of town lynches.

Even if I calculate it like this, it's 90 instead of 92%. If we know the amount of bussing, I can give the accurate percentage.

I literally have no idea what words you are saying.
If 8% of Town lynches have no Mafia on them, the other 92% have at least one. Am I crazy?

[goodmorning]

Anyway, there definitely is a little wiggle room in those percentages because no-lynches, but it's not enough that I want to spend time fixing it and anyway couldn't because forced no-lynches.

[goodmorning]

And replacement discussion would be better off in Toomai's stats thread. I'm not into those stats, I was more curious about VCA and whether it's as good as everyone thinks.

EDIT: well, I am into those stats, it's just not in the scope of this project because it's been covered.

[TierShift]

No gm, that is 8% of total lynches, not just town lynches.

[Lowercase]

On an unrelated note, I'm surprised how often BP's get lynched. Seems like it would never be a good idea to lynch a BP without a counter.