Dominion Strategy Forum

Please login or register.

Login with username, password and session length
Pages: 1 ... 29 30 [31] 32 33 ... 65  All

Author Topic: M124: GPT-2 Mafia (Game over ; Mafia wins flawlessly)  (Read 174671 times)

0 Members and 1 Guest are viewing this topic.

2.71828.....

  • Saboteur
  • *****
  • Offline Offline
  • Posts: 1290
  • Shuffle iT Username: irrationalE
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #750 on: September 10, 2019, 11:38:42 am »

unless anyone thinks we are getting a lynch off before the duel they should keep their reads to a minimum for now imo.

that hurt my soul to say, but I stand by it.

I disagree with this. Stifling discussion is anti-town. We want people to make the most informed decision possible when faced with the duel. If everyone generally views certain people as town or scum, then they appear in the duel, it is a fairly big data point once we know the success/failure of the duel.

But also, the most informed decision includes weighing the huge benefits of getting a PR out of the duel
Logged
Man. I had four strips of bacon yesterday. Was one automatically undercooked, one automatically overcooked? No, let's put a stop to that right here, all four strips were excellent.

Joseph2302

  • Jester
  • *****
  • Offline Offline
  • Posts: 858
  • Shuffle iT Username: Joseph2302
  • "Better to be lucky than good"
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #751 on: September 10, 2019, 11:53:42 am »

unless anyone thinks we are getting a lynch off before the duel they should keep their reads to a minimum for now imo.

that hurt my soul to say, but I stand by it.

I disagree with this. Stifling discussion is anti-town. We want people to make the most informed decision possible when faced with the duel. If everyone generally views certain people as town or scum, then they appear in the duel, it is a fairly big data point once we know the success/failure of the duel.
True but on the other hand, the more we say, the more scum gets ideas what we might vote, and thus what's their safest vote claim
Logged
Mafia Stats: (correct as of 2017)
Town: 22 games, 8 wins
Scum: 5 games, 3 wins

Joseph2302

  • Jester
  • *****
  • Offline Offline
  • Posts: 858
  • Shuffle iT Username: Joseph2302
  • "Better to be lucky than good"
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #752 on: September 10, 2019, 11:54:12 am »

unless anyone thinks we are getting a lynch off before the duel they should keep their reads to a minimum for now imo.

that hurt my soul to say, but I stand by it.

I disagree with this. Stifling discussion is anti-town. We want people to make the most informed decision possible when faced with the duel. If everyone generally views certain people as town or scum, then they appear in the duel, it is a fairly big data point once we know the success/failure of the duel.

But also, the most informed decision includes weighing the huge benefits of getting a PR out of the duel
Agree with this though. Duels are good
Logged
Mafia Stats: (correct as of 2017)
Town: 22 games, 8 wins
Scum: 5 games, 3 wins

jotheonah

  • Jester
  • *****
  • Offline Offline
  • Posts: 989
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #753 on: September 10, 2019, 12:14:35 pm »

How long do we have until duel time?
Logged
"I know old meta, and joth is useless day 1 but awesome town day 3 and on." --Teproc

He/him

2.71828.....

  • Saboteur
  • *****
  • Offline Offline
  • Posts: 1290
  • Shuffle iT Username: irrationalE
    • View Profile
Re: M124: GPT-2 Mafia (Day 1)
« Reply #754 on: September 10, 2019, 12:31:13 pm »

Phase 1 of day 2 starts now and ends at September 11, 05:00 Forum Time.
Logged
Man. I had four strips of bacon yesterday. Was one automatically undercooked, one automatically overcooked? No, let's put a stop to that right here, all four strips were excellent.

Joseph2302

  • Jester
  • *****
  • Offline Offline
  • Posts: 858
  • Shuffle iT Username: Joseph2302
  • "Better to be lucky than good"
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #755 on: September 10, 2019, 12:42:42 pm »

So just over 16 hours
Logged
Mafia Stats: (correct as of 2017)
Town: 22 games, 8 wins
Scum: 5 games, 3 wins

A Drowned Kernel

  • 2015 World Champion
  • *
  • Offline Offline
  • Posts: 1067
  • They/Them
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #756 on: September 10, 2019, 02:34:09 pm »

unless anyone thinks we are getting a lynch off before the duel they should keep their reads to a minimum for now imo.

that hurt my soul to say, but I stand by it.

I disagree with this. Stifling discussion is anti-town. We want people to make the most informed decision possible when faced with the duel. If everyone generally views certain people as town or scum, then they appear in the duel, it is a fairly big data point once we know the success/failure of the duel.

But also, the most informed decision includes weighing the huge benefits of getting a PR out of the duel

I think "huge" is an overstatement
Logged
The perfect engine
But it will never go off
Three piles are empty

A Drowned Kernel

  • 2015 World Champion
  • *
  • Offline Offline
  • Posts: 1067
  • They/Them
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #757 on: September 10, 2019, 02:35:32 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote
Logged
The perfect engine
But it will never go off
Three piles are empty

Swowl

  • Pawn
  • **
  • Offline Offline
  • Posts: 1
  • Shuffle iT Username: Swowl
  • Snow.Owl.Duh.
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #758 on: September 10, 2019, 02:42:27 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

This ^^
Is what i was going for
Logged
Jesus how did i fuck that up
real good. That’s how
Town:  14 wins, 14 losses (1 MVP)
Skum: 7 wins, 7 losses (1 MVP)
3-Party: 4 wins, 1 loss

EFHW

  • Spy
  • ****
  • Offline Offline
  • Posts: 81
  • Shuffle iT Username: EFHW
  • EFHW="ee-foo". Really, how else would you say it?
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #759 on: September 10, 2019, 03:01:52 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

This ^^
Is what i was going for
Ok, that seems fine.
Logged

EFHW

  • Spy
  • ****
  • Offline Offline
  • Posts: 81
  • Shuffle iT Username: EFHW
  • EFHW="ee-foo". Really, how else would you say it?
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #760 on: September 10, 2019, 03:04:40 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

This ^^
Is what i was going for
Ok, that seems fine.
I see e was concerned about stifling conversation. That is a consideration, but we want scum to be in the duel. So anything that increases that likelihood seems like a decent idea.
Logged

Debatepro

  • Pawn
  • **
  • Offline Offline
  • Posts: 2
  • Shuffle iT Username: Debatepr0
    • View Profile
Re: M124: GPT-2 Mafia (Day 1)
« Reply #761 on: September 10, 2019, 03:23:54 pm »

I’m camping I’m the worst, I’ll catch up and try harder

Still camping?
Logged
FNG

Robz888

  • Margrave
  • *****
  • Offline Offline
  • Posts: 2644
  • Shuffle iT Username: Robz888
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #762 on: September 10, 2019, 03:32:34 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

We get around this if we all just pledge to vote for the duel, and anyone who doesn't is considered automatically scum.
Logged
I have been forced to accept that lackluster play is a town tell for you.

Glooble

  • Bishop
  • ****
  • Offline Offline
  • Posts: 113
    • View Profile
    • Solutions to Problems
Re: M124: GPT-2 Mafia (Day 2)
« Reply #763 on: September 10, 2019, 03:54:44 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

We get around this if we all just pledge to vote for the duel, and anyone who doesn't is considered automatically scum.

It's so tempting isn't it? Anyone wanna redo that day 1 math accounting for fewer scum?
Logged

I think town!Glooble pointing to something as a scum tell and then shortly thereafter doing that thing is a lot more likely than scum!Glooble doing that.

He/ Him

Check out my podcast: www.stppodcast.com

Glooble

  • Bishop
  • ****
  • Offline Offline
  • Posts: 113
    • View Profile
    • Solutions to Problems
Re: M124: GPT-2 Mafia (Day 2)
« Reply #764 on: September 10, 2019, 04:03:19 pm »

I'm getting a 54.5% chance of town vs. town (9/12 for the first selection, 8/11 for the second.) Feel free to double check me.

So pledging to vote for the duel no matter what still seems like a risky plan.

The power role is slightly more useful now, but its still not a slam dunk.
Logged

I think town!Glooble pointing to something as a scum tell and then shortly thereafter doing that thing is a lot more likely than scum!Glooble doing that.

He/ Him

Check out my podcast: www.stppodcast.com

jotheonah

  • Jester
  • *****
  • Offline Offline
  • Posts: 989
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #765 on: September 10, 2019, 04:23:09 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

We get around this if we all just pledge to vote for the duel, and anyone who doesn't is considered automatically scum.

It's so tempting isn't it? Anyone wanna redo that day 1 math accounting for fewer scum?

you mean for fewer town, right?
Logged
"I know old meta, and joth is useless day 1 but awesome town day 3 and on." --Teproc

He/him

jotheonah

  • Jester
  • *****
  • Offline Offline
  • Posts: 989
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #766 on: September 10, 2019, 04:24:41 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

We get around this if we all just pledge to vote for the duel, and anyone who doesn't is considered automatically scum.

Scum will still vote how they want and just lie. And if the duel fails, it will only tell us that we failed to get enough town on board with the plan.
Logged
"I know old meta, and joth is useless day 1 but awesome town day 3 and on." --Teproc

He/him

Glooble

  • Bishop
  • ****
  • Offline Offline
  • Posts: 113
    • View Profile
    • Solutions to Problems
Re: M124: GPT-2 Mafia (Day 2)
« Reply #767 on: September 10, 2019, 04:48:46 pm »

The argument I'm considering is that in the event that the proposed duel ends up being town vs scum, we want scum to have the least clear idea of their chances of winning. We can always have discussion after the vote

We get around this if we all just pledge to vote for the duel, and anyone who doesn't is considered automatically scum.

It's so tempting isn't it? Anyone wanna redo that day 1 math accounting for fewer scum?

you mean for fewer town, right?

Yes.
Logged

I think town!Glooble pointing to something as a scum tell and then shortly thereafter doing that thing is a lot more likely than scum!Glooble doing that.

He/ Him

Check out my podcast: www.stppodcast.com

Swowl

  • Pawn
  • **
  • Offline Offline
  • Posts: 1
  • Shuffle iT Username: Swowl
  • Snow.Owl.Duh.
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #768 on: September 10, 2019, 05:02:47 pm »

the math is slightly less relevant now that Day 1 is past and we have some flips. or I mean... not less relevant, but the important factor is simply "how likely is skum to be in the duel?"

Which is slightly over 50% if my head math is correct.
Logged
Jesus how did i fuck that up
real good. That’s how
Town:  14 wins, 14 losses (1 MVP)
Skum: 7 wins, 7 losses (1 MVP)
3-Party: 4 wins, 1 loss

A Drowned Kernel

  • 2015 World Champion
  • *
  • Offline Offline
  • Posts: 1067
  • They/Them
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #769 on: September 10, 2019, 07:10:24 pm »

the math is slightly less relevant now that Day 1 is past and we have some flips. or I mean... not less relevant, but the important factor is simply "how likely is skum to be in the duel?"

Which is slightly over 50% if my head math is correct.

The math is completely irrelevant because now people should actually have reads to base their votes on
Logged
The perfect engine
But it will never go off
Three piles are empty

jotheonah

  • Jester
  • *****
  • Offline Offline
  • Posts: 989
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #770 on: September 10, 2019, 08:01:40 pm »

That’s silly. Reads should cause you to adjust your priors, but they certainly don’t make the math irrelevant.
Logged
"I know old meta, and joth is useless day 1 but awesome town day 3 and on." --Teproc

He/him

mcmcsalot

  • Explorer
  • *****
  • Offline Offline
  • Posts: 338
  • Shuffle iT Username: mcmcsalot
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #771 on: September 10, 2019, 08:26:06 pm »

Okay I am caught up.

I don’t think we should force any duel vote but I do think the pr is even more important and there is a greater chance pair contains at least one scum read.
Logged
Wins: M19, M21, M23, M24, M26, M39, M91, M94, M102, M104, M107, M114, M115
Losses: M20, M22, M25, M27, M30, M31, M35, M38, M40, M42, M46, M60, M90, M93, M96, M98, M100, M101, M106, M111, M113
Winrate: 38.2%(13/34) 29.6%(8/27), 71.4%(5/7)

mcmcsalot

  • Explorer
  • *****
  • Offline Offline
  • Posts: 338
  • Shuffle iT Username: mcmcsalot
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #772 on: September 10, 2019, 08:26:41 pm »

Joth v glooble looks pretty town v town. But I’m also not sold on either of the cases.
Logged
Wins: M19, M21, M23, M24, M26, M39, M91, M94, M102, M104, M107, M114, M115
Losses: M20, M22, M25, M27, M30, M31, M35, M38, M40, M42, M46, M60, M90, M93, M96, M98, M100, M101, M106, M111, M113
Winrate: 38.2%(13/34) 29.6%(8/27), 71.4%(5/7)

Robz888

  • Margrave
  • *****
  • Offline Offline
  • Posts: 2644
  • Shuffle iT Username: Robz888
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #773 on: September 10, 2019, 11:22:47 pm »

Joth v glooble looks pretty town v town. But I’m also not sold on either of the cases.

I could also see, in some weird way, Joth v Glooble as scum v scum. It would explain some of the "this does not make sense"-ness of it. But yeah most likely they are just both town.
Logged
I have been forced to accept that lackluster play is a town tell for you.

Swowl

  • Pawn
  • **
  • Offline Offline
  • Posts: 1
  • Shuffle iT Username: Swowl
  • Snow.Owl.Duh.
    • View Profile
Re: M124: GPT-2 Mafia (Day 2)
« Reply #774 on: September 11, 2019, 01:12:42 am »

Throwing this out there in regards to the theory of making people vote mandatory, but also being able to isolate if someone lies better than just "someone lied". No idea if it actually can be used as utility or not... someone with a less tired brain should read it over.

Here is the idea:

1) We group players randomly based on the amount of votes YES needed to get a duel to go. Since ties are YES, and we have 12, today that would be groups of 6.

2) We then take something arbitrary, such as the sign up list, to assign an order for the grouping process.

3) With the sign up list being used, it would be something like this:
1 - votes YES to 2-7
2 - votes YES to 3-8
3 - votes YES to 4-9
4 - votes YES to 5-10
5 - votes YES to 6-11
6 - votes YES to 7-12
7 - votes YES to 8-1
8 - votes YES to 9-2
9 - votes YES to 10-3
10 - votes YES to 11-4
11 - votes YES to 12-5
12 - votes YES to 1-6


4) Then I guess it would have to be something like - "Only vote yes if you are supposed to, if your dueled player is the lowest numeral"
   - So if you are supposed to vote yes for Player 6, but vote no for Player 4... you would vote no. This just ensures the grouping is kept to a minimum.

As I mentioned, it doesn't really have a lot of utility purpose. It pretty much makes all reads obsolete because we would have to agree to it prior to getting to the duel. All it really does is make sure the duel happens, or if it doesn't, it gives us a smaller pool of players to look at. This is to say:

- Player 1 and Player 7 are put at bat.
- Players 7-12 should vote YES, since 1 is the lowest numeral and on their list.
- All other players, including the players who are YES votes for Player 7, would vote NO.

At the end we either gain the duel (which could be good) or we gain the info that someone did not follow the rules... but only on the NO voting side... which is important. In the scenario above where we come to and there is no duel, we would learn that at least one player within Players 7-12 are skum (1/6).


No super like... pushing this plan rn... just more throwing it out there.
Logged
Jesus how did i fuck that up
real good. That’s how
Town:  14 wins, 14 losses (1 MVP)
Skum: 7 wins, 7 losses (1 MVP)
3-Party: 4 wins, 1 loss
Pages: 1 ... 29 30 [31] 32 33 ... 65  All
 

Page created in 0.129 seconds with 19 queries.