The Newbie Matrix6 stats thread (complete)

[Cheery Dog]

I think as a whole we catch more experienced scum, because those players would normally be early nightkills, although this is pretty much a random guess as to a reason.

New scum are either a hit or miss depending on how much experience they've had before and their understanding of the game.

[Faraday]

Do you have the setups ran for each? (I Presume random chance didn't dictate the 2 see setup got all the mafia sided games, but did it get more/less of them?)

[TierShift]

Setups probably is not going to be a confounder.

[Faraday]

Given the large difference it seemed worth checking if he has the stats at hand.

[quadz08]

I do want to point out that once broken out into newbie/se/ic configurations, most of the subtypes have fairly small sample sizes (<20 total games).

[Zachrulez]

I do want to point out that once broken out into newbie/se/ic configurations, most of the subtypes have fairly small sample sizes (<20 total games).

To get reasonable sample sizes for some of the subtypes, you would have to play an incredibly large number of newbie games under the setup due to the unlikelihood of some of the draws happening randomly. At some point you do need to theorize over having hard data because it's not feasable you'll ever have hard data on some of it.

Anyway because I'm somewhat bored and because I do believe that players will approach experience vs newbies in a different way, I think it's worthwhile to point out based on the data available that at least one IC or SE drew scum in the 3 SE, 1 IC setup approx 71.5% of the time. With 2 SE, 1 IC, the draw for at least one was approx 62.6%. I'm sure there's a way to figure out the mathematical probability, but that's for someone way more crazy into the math than I am.

Anyway, if I'm one of the SEs or ICs and I happen to draw town, I'm sure a compelling argument can be made toward lynching one of the more experienced players based entirely on the probability of one of them being scum alone. (I just pick the least likely to be town and invent some reason to lynch them.)

I'm sure the validity of that approach could be called into question, but my thought is to whether or not those kind of factors should be mitigated by reducing the presence of experience vs newbies in a newbie game. My thought is there should be just enough to teach the newbies how to play and no more.

Maybe 2 ICs is not sustainable, but would the notion of going 1 SE, 1 IC be a totally ludicrous idea to consider? That would actually bring us pretty close to the 2 IC ideal we had way back when, but with 1 IC actively expected to teach the game and 1 SE to offer alternative advice or to call out the SE if they're scum bsing? Is it important to have a third experience role? Does one of them have to be guaranteed to be town? (In 1 SE, 1 IC they could both be scum, but it would be rare.)

[Toomai]

I'm sure there's a way to figure out the mathematical probability

Assuming pure randomness, the odds of any particular type of scumteam are:

• 5Nb-3SE-1IC
  • Nb-Nb: 20/72 (~27.8%)
  • Nb-SE: 30/72 (~41.7%)
  • Nb-IC: 10/72 (~13.9%)
  • SE-SE: 6/72 (~8.3%)
  • SE-IC: 6/72 (~8.3%)
• 6Nb-2SE-1IC
  • Nb-Nb: 30/72 (~41.7%)
  • Nb-SE: 24/72 (~33.3%)
  • Nb-IC: 12/72 (~16.7%)
  • SE-SE: 2/72 (~2.8%)
  • SE-IC: 4/72 (~5.6%)

[Cogito Ergo Sum]

Anyway, if I'm one of the SEs or ICs and I happen to draw town, I'm sure a compelling argument can be made toward lynching one of the more experienced players based entirely on the probability of one of them being scum alone. (I just pick the least likely to be town and invent some reason to lynch them.)

The odds of there being scum among the other non-newbies is literally the same odds as the odds of there being scum among any group of players in that game of the same size. There's no substance to this argument.

[Zachrulez]

Anyway, if I'm one of the SEs or ICs and I happen to draw town, I'm sure a compelling argument can be made toward lynching one of the more experienced players based entirely on the probability of one of them being scum alone. (I just pick the least likely to be town and invent some reason to lynch them.)

The odds of there being scum among the other non-newbies is literally the same odds as the odds of there being scum among any group of players in that game of the same size. There's no substance to this argument.

I'm not arguing that it's optimal play necessarily, just pointing out that those kind of stats can have power in an environment where experienced players can be seen as bigger potential threats.

[Toomai]

For the record, the one remaining scum of Newbie 1574 conceded the game. To keep the stats functional, I have recorded this as a lynch on the offending scumbum.

[goodmorning]

Dammit, that's going to bother mine.

[Toomai]

Okay this is officially ridiculous.

The slot replacement rate for newbie scum slots is now 87.5%, and the player replacement rate is now 110.6%.

In other words, newbie scum slots have a 7/8 chance of seeing at least one replacement, with an average replacements-per-game of 1.1.

And because I don't count most pregame replacements, nor replacements that are another person's fault, this is a lower bound.

[BBmolla]

Newbie town slots?

[Cheery Dog]

If we gave scum daytalk, I think that would help decrease the amount of newbie scum replacements.

[~Jordan`]

daytalk is the best

[Toomai]

Newbie town slots?

...get replaced less than half as often. (42% of them saw at least one replacement, average of 0.515 replacements per game)

If we gave scum daytalk, I think that would help decrease the amount of newbie scum replacements.

This is true, but scum already has a winning percentage advantage, and it's not necessarily representative of the site as a whole.

[Oversoul]

Wow. That is extreme.

[Not_Mafia]

So any newbie slot with a replacement has ~37% of being scum, which is 68% more accurate than random lynching, not ideal but it doesn't seem too bad, at least not as bad as the 87.5% looks initally

[Toomai]

Okay this is gonna make me look stupid but here it is.

I found a major bug in the replacements pivot tables. Basically, all slots in all games since 1531 were being counted as both town and scum for the question of "how many times did this slot get replaced". I don't have a clue how it happened, nor why Excel didn't pop up one of its "these adjacent formulas don't match" boxes to tell me.

So expect some major changes to the replacement stats in the next little bit. I suspect things will look far less ridiculous once it's all corrected.

[goodmorning]

That explains a lot.

[JeanDarc]

I feel that the forum titles should have some dark humour.

1. anoints the newbies who replace OUT of their first two games or two games in a row (prior to completion of any of their games) with the title: Weasel; or Ferret
2. commend other players who replace INTO x games in a row and stick to end of game/death with the title: Tree Stump; or better yet, Master Disguiser.

I wonder how many games would suffice for the honourable title of a Tree Stump, but I think 10 is a good start.

[Toomai]

Okay so this is somewhat interesting: setup A (JK+RB+BP) has gone on a town win streak lately and is now above 40% like all the other subsetups. It remains the lest-common setup too, so there's still room to improve as well.

[Runner]

Wow these stats are very... thorough. Somewhat painful to read how one game lasted nearly 3 months, and how one game had 11 players replace out.

Thanks for putting in the time to do this stuff Toomai. Do you think the below average winrate for setup A will rise once the number of games completed for that setup rises more in line with the rest?

[Toomai]

Do you think the below average winrate for setup A will rise once the number of games completed for that setup rises more in line with the rest?

30 samples is traditionally where statisticians put the "minimum" mark for a binary collection. Since all the subsetups have now passed that, I don't expect the winrates to change too much from now on unless one faction's meta gets a power boost.

[Toomai]

I have begun to add the stats for the ongoing daytalk experiment to the first post. With only 2 completed games so far, there's nothing important to say about it so far other than it's there.