Dominion Strategy Forum

Please login or register.

Login with username, password and session length
Pages: 1 ... 3 4 [5]  All

Author Topic: IsoHearth I: Nkirbit Wins!  (Read 11060 times)

0 Members and 1 Guest are viewing this topic.

HiveMindEmulator

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2222
  • Respect: +2079
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #100 on: February 09, 2015, 12:18:48 pm »
0

Rule clarification questions:
1. Are we supposed to reveal what our 3 classes are before the match or not?
2. What is the timeframe for the "rounds" and does the order matter?
« Last Edit: February 09, 2015, 12:20:35 pm by HiveMindEmulator »
Logged

popsofctown

  • Governor
  • *****
  • Offline Offline
  • Posts: 4996
  • Respect: +2520
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #101 on: February 09, 2015, 12:37:50 pm »
0

Supose more probable win distribution. Here's results for both round robin and double elimination (calculated, not simulated):

Win probabilities
1   0.0   0.6   0.7   0.8
2   0.4   0.0   0.6   0.7
3   0.3   0.4   0.0   0.6
4   0.2   0.3   0.4   0.0
Round robin
Advance
1 0.7651199999999999
2 0.59824
3 0.40175999999999995
4 0.23488
Win
1 0.47312
2 0.28424
3 0.16176000000000004
4 0.08088
Double elimination
Advance
1 0.7858400000000001
2 0.6075466666666668
3 0.39245333333333327
4 0.21416000000000004
Win
1 0.5531754666666664
2 0.2763130666666666
3 0.1245696
4 0.04594186666666668

So both if we need to determine 2 advances and if we need to determine single winner double elimination is better. In case you provided there is no difference between double elimination and round robin - only the fact that however was paired with trash was decided by draw and not by skill, but supposedly skills are same

I'm not sure how you judge that double elimination is better based on that. The winner distribution is lower entropy because more games are played, and by averaging over all potential initial brackets you avoided the major issue which is the impact of the initial matchup. The problem is that the result depends too much on who plays whom first.

Player one is the best player and also wins more of the time in double elimination.  I haven't checked his math yet, though.
Logged
Also you probably are an expert if you buy two bureaucrats early.

HiveMindEmulator

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2222
  • Respect: +2079
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #102 on: February 09, 2015, 12:52:37 pm »
0

This is a single example of a win probability matrix, and it's clear that player 1 is the best because he is 50%+ in all matchups.
So any reasonable format should have him win more of the time, and the one that involves more games will increase that chance. This is not particularly interesting.

In fact, let's make this super mathy...
Let the win probability matrix be a random variable X, and call the max wieght row k
Let the format be a mapping f:X-->y, where y is the win probability vector.

In this example it seems we are trying to maximize y(k) subject to X=A.

But I contend this is not what we want to maximize. What we want is that the random winner be the best predictor of the best player, that is:
maximize P(A(i) >= A(j) for all rows j | z = i, where z is drawn from the distribution given by y).

By making A non-random, this example doesn't really show anything.
Logged

popsofctown

  • Governor
  • *****
  • Offline Offline
  • Posts: 4996
  • Respect: +2520
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #103 on: February 09, 2015, 03:01:56 pm »
0

This is a single example of a win probability matrix, and it's clear that player 1 is the best because he is 50%+ in all matchups.
So any reasonable format should have him win more of the time, and the one that involves more games will increase that chance. This is not particularly interesting.

In fact, let's make this super mathy...
Let the win probability matrix be a random variable X, and call the max wieght row k
Let the format be a mapping f:X-->y, where y is the win probability vector.

In this example it seems we are trying to maximize y(k) subject to X=A.

But I contend this is not what we want to maximize. What we want is that the random winner be the best predictor of the best player, that is:
maximize P(A(i) >= A(j) for all rows j | z = i, where z is drawn from the distribution given by y).

By making A non-random, this example doesn't really show anything.
Wait, a double elim bracket takes more games than round robin, for four people?
Logged
Also you probably are an expert if you buy two bureaucrats early.

HiveMindEmulator

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2222
  • Respect: +2079
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #104 on: February 09, 2015, 03:10:29 pm »
0

Yeah. Round Robin is 6 games. Double elim is 6 + P(LB winner beats WB winner)
Logged

nkirbit

  • Conspirator
  • ****
  • Offline Offline
  • Posts: 238
  • Respect: +45
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #105 on: February 09, 2015, 04:25:24 pm »
0

Nkirbit beats Twistedarcher 3-2.
Logged

nkirbit

  • Conspirator
  • ****
  • Offline Offline
  • Posts: 238
  • Respect: +45
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #106 on: February 09, 2015, 05:02:38 pm »
0

nkirbit 3 - HME 0.
Logged

popsofctown

  • Governor
  • *****
  • Offline Offline
  • Posts: 4996
  • Respect: +2520
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #107 on: February 09, 2015, 05:29:53 pm »
0

Kirbyhero over twisted archer 3-2
Logged
Also you probably are an expert if you buy two bureaucrats early.

popsofctown

  • Governor
  • *****
  • Offline Offline
  • Posts: 4996
  • Respect: +2520
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #108 on: February 09, 2015, 05:32:19 pm »
0

i tried to stream and record the set against twisted archer but the graphics froze and i only got audio
Logged
Also you probably are an expert if you buy two bureaucrats early.

ashersky

  • Board Moderator
  • *
  • Offline Offline
  • Posts: 2312
  • 2013/2014/2015 Mafia Mod of the Year
  • Respect: +1492
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #109 on: February 09, 2015, 06:18:16 pm »
0

Rule clarification questions:
1. Are we supposed to reveal what our 3 classes are before the match or not?
2. What is the timeframe for the "rounds" and does the order matter?

1.  Not required, but allowed.
2.  I would generally say 2 days per round.  Since the format is round robin, the only reason to force the rounds to be played in order (i.e., no round 2 matches until round 1 is over) would be to build suspense.  So I think you can just play your matches.
Logged
f.ds Mafia Board Moderator

2013, 2014, 2015 Mafia Mod of the Year

Games (30): Mafia: 17, 21, 24, 31, 39, 43, 67, 72, 86 || Newbie: 7 || RMM: 6, 7, 10, 20, 22, 26, 32, 45 || Blitz: 4-7, X, 17-19, 22, 24 || BM: 14, 19

2015 f.ds Representative, World Forum Mafia Championships
2013, 2014 Mafia Player of the Year (Tie)

Town: 46 Games, 23 Wins
Scum: 21 Games, 12 Wins
10x MVP: M30, M83, ZM16, M25, M38, M61, M76, RMM5, RMM41, M51

EgorK

  • Apprentice
  • *****
  • Offline Offline
  • Posts: 272
  • Respect: +73
    • View Profile
Re: IsoHearth I: Starting Tomorrow
« Reply #110 on: February 10, 2015, 02:04:19 am »
0

Supose more probable win distribution. Here's results for both round robin and double elimination (calculated, not simulated):

Win probabilities
1   0.0   0.6   0.7   0.8
2   0.4   0.0   0.6   0.7
3   0.3   0.4   0.0   0.6
4   0.2   0.3   0.4   0.0
Round robin
Advance
1 0.7651199999999999
2 0.59824
3 0.40175999999999995
4 0.23488
Win
1 0.47312
2 0.28424
3 0.16176000000000004
4 0.08088
Double elimination
Advance
1 0.7858400000000001
2 0.6075466666666668
3 0.39245333333333327
4 0.21416000000000004
Win
1 0.5531754666666664
2 0.2763130666666666
3 0.1245696
4 0.04594186666666668

So both if we need to determine 2 advances and if we need to determine single winner double elimination is better. In case you provided there is no difference between double elimination and round robin - only the fact that however was paired with trash was decided by draw and not by skill, but supposedly skills are same

I'm not sure how you judge that double elimination is better based on that. The winner distribution is lower entropy because more games are played, and by averaging over all potential initial brackets you avoided the major issue which is the impact of the initial matchup. The problem is that the result depends too much on who plays whom first.

I was more interested in advancement percantages. And there double elim is supposedly better even with less game played. Also if there is no clear order I do not see any metrics to use
Logged

ashersky

  • Board Moderator
  • *
  • Offline Offline
  • Posts: 2312
  • 2013/2014/2015 Mafia Mod of the Year
  • Respect: +1492
    • View Profile
Re: IsoHearth I: Round 2
« Reply #111 on: February 10, 2015, 03:39:46 am »
0

Round 1 results:

nkirbit (Nick) defeats Twistedarcher (bkirbit)
mikohoy (HME) defeats kirbyhero (pops)

Round 2 results:

nkirbit (Nick) defeats HME (mikohoy)
kirbyhero (pops) defeats TwistedArcher (bkirbit)

Current standings:

nkirbit (2 points)
HME (1 point)
pops (1 point)
TA (0 points)

Round 3 remains:

Round 3:

TA vs. HME
nkirbit vs. Pops
Logged
f.ds Mafia Board Moderator

2013, 2014, 2015 Mafia Mod of the Year

Games (30): Mafia: 17, 21, 24, 31, 39, 43, 67, 72, 86 || Newbie: 7 || RMM: 6, 7, 10, 20, 22, 26, 32, 45 || Blitz: 4-7, X, 17-19, 22, 24 || BM: 14, 19

2015 f.ds Representative, World Forum Mafia Championships
2013, 2014 Mafia Player of the Year (Tie)

Town: 46 Games, 23 Wins
Scum: 21 Games, 12 Wins
10x MVP: M30, M83, ZM16, M25, M38, M61, M76, RMM5, RMM41, M51

nkirbit

  • Conspirator
  • ****
  • Offline Offline
  • Posts: 238
  • Respect: +45
    • View Profile
Re: IsoHearth I: Round 3
« Reply #112 on: February 10, 2015, 07:39:04 pm »
0

nkirbit 3 - pops 2
Logged

popsofctown

  • Governor
  • *****
  • Offline Offline
  • Posts: 4996
  • Respect: +2520
    • View Profile
Re: IsoHearth I: Round 3
« Reply #113 on: February 11, 2015, 11:45:06 am »
0

Nick takes the tourney!
Logged
Also you probably are an expert if you buy two bureaucrats early.

HiveMindEmulator

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2222
  • Respect: +2079
    • View Profile
Re: IsoHearth I: Nkirbit Wins!
« Reply #114 on: February 12, 2015, 02:07:23 pm »
0

mikohoy 3 - bkirbit 2

Now that we're all done, are deck lists going to be posted? It would be interesting if everyone knew how the deck matchups went.

This one was
me : TA
warlock < shaman
paladin > warrior
warlock < warrior
priest > hunter
warlock > hunter

vs nick was:
warlock < warrior
paladin < mage
paladin < hunter

vs pops I don't quite remember, but I think it was:
warlock > priest
paladin > warrior
priest > priest
Logged

nkirbit

  • Conspirator
  • ****
  • Offline Offline
  • Posts: 238
  • Respect: +45
    • View Profile
Re: IsoHearth I: Nkirbit Wins!
« Reply #115 on: February 12, 2015, 02:31:04 pm »
0

Mine (I think) were:

Me : TA

Warrior > Warrior
Mage > Warrior
Hunter < Warrior
Hunter < Hunter
Hunter > Shaman

vs HME was:
warlock < warrior
paladin < mage
paladin < hunter

vs Pops was:

hunter < Warrior
Warrior > Priest
mage < priest
mage > paladin
hunter > paladin

I'm not 100% sure about the match vs pops.  I know that the order of wins and my classes were right, but his might be off.

I definitely wouldn't play my hunter deck again.  I ended up going 3-3 with it, and all of the losses were not at all close and all of the wins I barely got and I felt like I drew well above average.  I was very happy with my warrior deck, and my mage deck was okay.
Logged

popsofctown

  • Governor
  • *****
  • Offline Offline
  • Posts: 4996
  • Respect: +2520
    • View Profile
Re: IsoHearth I: Nkirbit Wins!
« Reply #116 on: February 12, 2015, 05:26:18 pm »
0

I was playing Druid, not Paladin
Logged
Also you probably are an expert if you buy two bureaucrats early.
Pages: 1 ... 3 4 [5]  All
 

Page created in 0.08 seconds with 21 queries.