Newbie Student Mafia XXIV - Page 73

Post a Reply

Prev 1 71 72 73

NeverUnlucky

Canada1622 Posts

November 08 2016 03:11 GMT

#1441

On November 08 2016 11:49 NocturneMage wrote:

Show nested quote +

it was a pleasure driving you up the wall.... xD

Show nested quote +

you have inspired me to be quite the spammer scumformation xD

Now that. That was quite a read.

ExO_

United States2316 Posts

November 08 2016 09:32 GMT

#1442

....this game ended rather abruptly. cannot believe skynx was mafia.

gg boys.

sorry about getting all upset rels

disformation

Germany8352 Posts

November 08 2016 09:49 GMT

#1443

On November 08 2016 12:11 NeverUnlucky wrote:

Show nested quote +

Now that. That was quite a read.

Capsformation.

beentheredonethat

2934 Posts

November 08 2016 12:30 GMT

#1444

cakepie could've hosted this alone, given that my contribution was basically 1 (one) single votecount. I feel sorry but everytime I checked (which was once to max. twice per real life day) there was nothing or very little to do.

cakepie

985 Posts

November 08 2016 18:04 GMT

#1445

Postgame Notes

Sorry for the delay. Had to go to bed after wrapping things up last night; been a long day today.
Some stuff in here rehashes things you've probably already read in Obs.

Thanks

Thanks to btdt for cohosting. I guess our timings ended up overlapping a lot -- but don't worry about it, I do appreciate having backup, it's important in case of contingencies.
Thanks to coaches Artanis, Shapelog, Half the Sky & Holyflare for volunteering their time to mentor.
Shoutouts to ptmc for being press-ganged into standing by as replacement, and Acrofales for being standby coach.
Thanks to players for participating.

Behavior / Toxicity

Mafia is a game that players can get very invested in, and get emotional about. It can also get pretty heated at times. Different hosts may differ slightly in terms of what they are willing to tolerate, but in general our goal is to have a community where we can (within certain limits) argue with one another within the bounds of the game, and then GG and be pals outside of the game.

My personal rules are best summarized as:

Heat-of-the-moment anger -- within the context of an intense moment in the game, i.e. EoD1, I'm usually willing to overlook an isolated outburst or brief heated exchange as long as it doesn't devolve into a protracted fight or recur later in the game. I believe we can all be mature and understanding enough that once taken out of the high-pressure situation, people will apologize, forgive and make good. If it develops into a personal vendetta, however, expect mod action.
Attacking the substance of a player's play -- logic, reasoning, playstyle -- is fine. Direct personal attacks are not.
Ad hominem is poor form, enough said.

OGI

I added the MafiaMetric experiment at the last minute and forgot to explicitly put in a rule against mentioning it within the game. I apologize for this omission, which opened up the opportunity for OGI.

I also had to make an announcement regarding relaxing the schedule for submitting reads. This was in fact due to me changing my mind for technical reasons and unrelated to any reads submitted before that point. I tried to make this announcement as NAI as possible without confirming nor denying the townslip by NeverUnlucky, but failed -- it remained open to interpretation that the change was triggered by an off-schedule read submission by him.

I'm sorry that the experiment was allowed to interfere with the game, I should have planned it more carefully rather than unleash it in the flash of inspiration.

I realize that this isn't the first time that NeverUnlucky has pulled something like this, but I would like to believe that it is not out of malicious intent, but over-eagerness to establish/confirm his town alignment, and ultimately, a drive to win the game. I have made very clear by PM to him that neither "establish yourself as town" nor "play to win" includes the use gamebreaking means. He has also already apologized to me by PM earlier today for the OGI and other matters.

Due to my own hosting error involvement, I will not seek to penalize NeverUnlucky for this, but will make a side note of it in the banlist submission for this game.

Banlist actions
See here.

MafiaMetric: graphs in a few days when I have more time to work on them.

mahrgell

Germany3943 Posts

November 08 2016 18:52 GMT

#1446

<3 cakepie
You will always be remember as my first! Nobody can take that from you!

cakepie

985 Posts

November 12 2016 05:52 GMT

#1447

Read Scoring
The score is an indicator of the accuracy of a submitted read and is measured as an error in terms of number of pairwise (mis)ordered players. The lower the number, the more accurate the read.

Each read submitted is evaluated by checking every possible pair of town+scum players. If the scum player is rated scummier than the town player, there is no penalty; if the order is reversed, there is a penalty of 1. If the two players are indistinguishable (same line) then the penalty is 0.5.

For an external observer with no extra information at the start of the game, there are 2x7=14 possible pairs to evaluate. On D2 there were 2x5=10 pairs. Divide the number of mismatches by the number of possible pairs to obtain a normalized value between zero and one.

A score of zero means that the read was perfectly correct, and a score of one means that it is completely wrong.
A random guesser will on average obtain a score of 0.5. A true null read that doesn't distinguish between any players (all names on one line) also scores 0.5.

disformation
Since disformation was such a obs champion, he gets his own scatter plot.
[image loading]

everybody
This is how read quality changed throughout the game.
This chart rates players/coaches on the same scale as obs, which is actually unfair (Players are confirmed town to themselves, so that's a couple of pairs that they're guaranteed to get correct for free. Obs do not have this information).
[image loading]

players only
This chart has just the players, and does not count the pairs involving the player themselves.
The error scores are thus slightly higher than the previous chart.
[image loading]

cakepie

985 Posts

November 12 2016 05:52 GMT

#1448

Scum Performance
Scum wants to avoid being lynched, and to do that, they need to avoid being scumread.

For each scum player, on each submitted read, we count the number of players that were considered scummier than them (again, indistinguishable counts as 0.5). This is normalized against the number of players alive.

A score of 0 means that the player was rated as the sole bottom scum, below everyone else in a submitted read.
A score of 1 means that the player was rated as sole top town, above everyone else.

Notes
This one is a bit wonky because the normalization is based on a total ordering -- i.e. every player on their own line -- but to be fair that's probably too much to ask for. The effect of reading some players as tied/indistinguishable are:
- If a few players are rated as tied for top town, they will score somewhat less than 1.
- If a couple of players are tied for bottom scum, they will score a bit more than 0.
(Mathematically speaking, N tied players are all assigned the median value of a band N players wide rather than each occupying one spot.)

In any case, a low score in the bottom 1/4 to 1/3 of players is not a good place to be.

Skynx
Notably, Shapelog had him pegged as bottomscum all game long. Took a dive in late D1 into N1, especially for Calix
[image loading]

Foreman
[image loading]

Overall
Looked okay at midday D1, but tended to be polarizing figures rather than blending into the middle of the pack. Stock tanked when town circle formed early D2.

cakepie

985 Posts

November 12 2016 05:52 GMT

#1449

Town Performance
The same charts as for scum players, but now for town players. In general higher = better.

The premise for this one is a little bit shakier, since this will depend on role and playstyle. Some players will gladly be strong toptown and die to scum N1, while others (and TPR) will want to survive for a bit in the middle of the pack.

Calix
disformation going tinfoil is very evident in this one, just look at those wild jumps! Otherwise pretty solidly upper half.
[image loading]

darthfoley
Data a bit sparse but looks like could have been an alternative wagon D1.
[image loading]

ExO_
Data sparse/incomplete but definitely a difficult hole to climb out of EoD1-- widely read as bottom 2 during second half of D1.
[image loading]

mahrgell
Sought death N1, succeeded in being the medic dodge.
[image loading]

NeverUnlucky
Seems like might have lost a few points for EoD1?
[image loading]

Rels
Chaotic.
[image loading]

Tictock
Inscrutable D1 into join town circle D2.
[image loading]

cakepie

985 Posts

November 12 2016 05:52 GMT

#1450

Thoughts

Short game, we didn't get to see longer progressions so limited conclusions.
Data was variously too dense (disformation!!) or too sparse. Drop off in submissions D2. Guideline of approx 8~12hr reporting interval might be ideal, with additional submissions for sudden/snap read changes.
Lacked submissions from some players, needs more buy-in from players to work better / more complete analysis.
Denser data is noisier? Or maybe it's just tinfoil!disfo? Not sure.
May be too much effort / distraction to submit reads, time consuming enough to just play, let alone do extra on the side.
Needs strict no-mention rules; even then, not sure if affects player behavior by making them more likely to revisit/reassess player ordering.

Feedback sought

What do you think is a good interval to report/submit at? How to strike a balance between minimizing player effort to submit reads vs getting useful output at the end?
Did you feel that it changed the way you played / processed your reads? E.g. did it force you to think about players that you would have otherwise set aside for the time being, etc.
Does this look like it might be useful, meaningful, or even just interesting fun? Would you do it again? Would you encourage others to participate?

154 Posts

November 12 2016 09:12 GMT

#1451

I can't seem to be able to express how deeply disappointed I am at this. These aren't pie charts. I was expecting delicious omnomno pie omnom charts.

Would you be interested in teaching me the shortcut of calculating bragons using microsoft excel or google spreadsheet power? Bannon was using some online timespan calculator, a standard calculator and good ol' pen and paper.

[(#Town alive - 1) / #Scum alive (including Serial Killer) × hours voting S(K)] - hours voting T - (0.2 per hour not voting)

Shapelog

United States5185 Posts

November 12 2016 23:54 GMT

#1452

I had to read it a few times, but do follow it. Will comment more in detail in a bit.

Shapelog

United States5185 Posts

November 14 2016 13:23 GMT

#1453

On November 12 2016 14:52 cakepie wrote:
Thoughts

Short game, we didn't get to see longer progressions so limited conclusions.
Data was variously too dense (disformation!!) or too sparse. Drop off in submissions D2. Guideline of approx 8~12hr reporting interval might be ideal, with additional submissions for sudden/snap read changes.
Lacked submissions from some players, needs more buy-in from players to work better / more complete analysis.
Denser data is noisier? Or maybe it's just tinfoil!disfo? Not sure.
May be too much effort / distraction to submit reads, time consuming enough to just play, let alone do extra on the side.
Needs strict no-mention rules; even then, not sure if affects player behavior by making them more likely to revisit/reassess player ordering.

Feedback sought

What do you think is a good interval to report/submit at? How to strike a balance between minimizing player effort to submit reads vs getting useful output at the end?
Did you feel that it changed the way you played / processed your reads? E.g. did it force you to think about players that you would have otherwise set aside for the time being, etc.
Does this look like it might be useful, meaningful, or even just interesting fun? Would you do it again? Would you encourage others to participate?

First thing first, I would flip the 0=you got scum team, 1=you are far off. Mostly due to the fact that 0 means off in binary, and 1 means on. Just something that is bothering me greatly.

In terms of the service.

I think you need to have submission times for this to work. Otherwise, people would spam the PM box with it. I also think you should have people sign up ahead of time, to see who WANTS to do it. That way, the people doing it will be more likely to submit their reads via PM.

I do not see it as extra work. Takes about 3-5 mins to copy and paste your reads into a PM and send them.

I would say the best Interval would be half the cycle time. Like we did before.
No, if anything. It just reinforced what I believed in. I feel some players would take that to the extreme, and hold bias to their initial reads. Butttttt, that happens anyways so :/
Its nice to have, and useful to have on file. But I really see it more as a bragging rights thing. Players should know how to improve just based on whom the scum team is postgame. These charts really don't give details into why they got X read, therefore, are not that great when it comes to having players improve. A post game analysis would be better.

Lunaticman

Sweden1097 Posts

November 18 2016 12:34 GMT

#1454

Amazing read, incredible hosting and impeccable coaches.

You are all so very awesome, TL is the place to be for mafia. Just saying.

Prev 1 71 72 73

Please or register to reply.

Newbie Student Mafia XXIV - Page 73

Completed

Ongoing

Upcoming