|
On April 12 2013 01:55 marvellosity wrote:
Just as an arbitrary example for this game, if we take 1500 ELO to be 'starting' ELO with 2500 being totally exceptional and 500 being very poor, you might see something like the following (apologise in advance for over/under estimating players, more for shits and giggles)
1. iamperfection; town ELO 2000 mafia ELO 1000
2. WaveofShadow; town ELO 1500 mafia ELO 1500
3. Ace; town ELO 2100 mafia ELO 2500
4. prplhz; town ELO 1600 mafia ELO 1200
5. DarthPunk; town ELO 2000 mafia ELO 2000
6. Oatsmaster; town ELO 1400 mafia ELO 1300
7. Tunkeg; town ELO 1500 mafia ELO 1500
8. Palmar; town ELO 2400 mafia ELO 2000
9. marvellosity; town ELO 2100 mafia ELO 2400
10. Axlegreaser; town ELO 1300 mafia ELO 1300
11. raynpelinkeet; town ELO 1600 mafia ELO 1600
12. sciberbia; town ELO 2100 mafia ELO 1700
Average strength of mafia team: 4100/3 = 1366.6
Average strength of town team: 17400/9 = 1933.3
Best theoretical mafia team: 6900/3 = 2300
Worst theoretical town team: 15000/9 = 1666.6
Average town ELO: 21600/12 = 1800
Average mafia ELO: 20000/12 = 1666.6
I'm not sure what the point of doing something like this is with no real mathematical way of estimating skill.
|
On April 12 2013 01:37 Oatsmaster wrote:Show nested quote +On April 12 2013 01:36 marvellosity wrote:On April 12 2013 01:29 wherebugsgo wrote:
I disagree still.
There is nothing that indicates a scumteam of sciberbia/DP/whoever would necessarily have been better than what we had.
You know why? Cause people like Axle, who play very differently dependent on alignment, are unlikely to get lynched as town.
It's all just conjecture, saying that you would have been able to balance the game better on that basis without considering those 3 vets. The argument is based on seeing the results, but there is no empirical evidence that it actually holds in practice (also there's no evidence that the reason the scumteam lost was because of player imbalance either!)
And to top it all off there's no measure of skill that says the scumteam you proposed is better than the one that was in this game, or that the change would have kept town level the same.
E: this was @ DP and whomever else was talking about the teams I'm kinda tempted, in a completely arbitrary manner, to do some kind of ELO rating for players with both a town and a mafia ELO, with results in games averaged across the team's ELO. Just for curiosity of how it looks. On a speculative note, you could say that you'd expect games with a higher proportion of players with large disparity between town and mafia ELOs to have more positive town results, regardless of 'objective' rating. For example a game of 12 iamperfections would more often result in a town win than a game of 12 wbgs. 12 Iamps. Hilarious
I thought it was very bright Idea.
|
United Kingdom36161 Posts
see my last post 
|
On April 12 2013 02:00 marvellosity wrote:Show nested quote +On April 12 2013 01:58 syllogism wrote:
While you are doing arbitrary math, can you factor in the advantage town gets from knowing that teams aren't RNGed Well the thing is, it becomes much less arbitrary if I go back many games and factor in results of each game, because ELO by its nature is self-correcting. So if I make a guess that's originally too high, if they lose games then their rating would fall quite quickly.
Yes but the point I'm making is, there is no REAL mathematical way of estimating skill like this. Any numbers you come up with or what a win or loss does to this arbitrary number is still just that, arbitrary.
As a player plays more and more games in a team setting, elo becomes more accurate supposedly, but with a sample size like TL Mafia, there is absolutely no way a skill rating like elo would come close to being an accurate representation. There are WAY too many mitigating factors such as the hosts, style of game (mini/normal/themed/whatever), roles available, who is playing in said game, etc. Theoretically this would matter less if people could play hundreds of games but that doesn't happen here.
|
On April 12 2013 02:01 AxleGreaser wrote:Show nested quote +On April 12 2013 01:37 Oatsmaster wrote:On April 12 2013 01:36 marvellosity wrote:On April 12 2013 01:29 wherebugsgo wrote:
I disagree still.
There is nothing that indicates a scumteam of sciberbia/DP/whoever would necessarily have been better than what we had.
You know why? Cause people like Axle, who play very differently dependent on alignment, are unlikely to get lynched as town.
It's all just conjecture, saying that you would have been able to balance the game better on that basis without considering those 3 vets. The argument is based on seeing the results, but there is no empirical evidence that it actually holds in practice (also there's no evidence that the reason the scumteam lost was because of player imbalance either!)
And to top it all off there's no measure of skill that says the scumteam you proposed is better than the one that was in this game, or that the change would have kept town level the same.
E: this was @ DP and whomever else was talking about the teams I'm kinda tempted, in a completely arbitrary manner, to do some kind of ELO rating for players with both a town and a mafia ELO, with results in games averaged across the team's ELO. Just for curiosity of how it looks. On a speculative note, you could say that you'd expect games with a higher proportion of players with large disparity between town and mafia ELOs to have more positive town results, regardless of 'objective' rating. For example a game of 12 iamperfections would more often result in a town win than a game of 12 wbgs. 12 Iamps. Hilarious I thought it was very bright Idea.
HAHAHAHAHA good one.
|
On April 12 2013 00:18 DarthPunk wrote:Show nested quote +On April 12 2013 00:10 marvellosity wrote:On April 12 2013 00:07 DarthPunk wrote:
To be fair. Ace/marv/palmar all being town wasn't even a major reason that town won this game. Vets often play badly just like everyone else. So if we are talking about balancing 'vets' I really don't see the point.
Just as a general point, it's more the idea that ace/marv/palmar can organise/direct a mafia team well, and can exert significant thread influence for the benefit of mafia. So could a whole bunch of non vets in this game. I still don't see the point. You know something I have noticed though is that 'town leaders' from newbie games really quiet down in the transition to normals, I think a lot of that has to do with all the vet nonsense and thus they are less likely to contribute or take charge as they have in newbies. I quite like the idea of different players taking charge, or at least having the chance to take charge of a scum team.Having 'good scum players' in every scum team would become stagnant and deprives newer players that opportunity to grow and shine. Even if they don't succeed the opportunity is important. Case in point. Look at oats in this game fucking take charge in the scum QT and improve DRASTICALLY from his last scum game. He rose to the occasion and even if he lost that still has a heap of value. Furthermore, Who doesn't love a good comeback? or the tale of an underdog overcoming the odds and succeeding? I think there are a lot of intangibles to consider that are more important than some easy wins every now and then.
agree with the bolded 10000000%. It's the main reason I take a backseat in many of the games I play. Let someone else develop and have a chance to take control of the game asTown or Scum. Some of my early game posts were just about keeping focus and avoidong what I thought were obvious mislynches. You guys did the heavy lifting.
|
On April 12 2013 02:04 WaveofShadow wrote:Show nested quote +On April 12 2013 02:00 marvellosity wrote:On April 12 2013 01:58 syllogism wrote:
While you are doing arbitrary math, can you factor in the advantage town gets from knowing that teams aren't RNGed Well the thing is, it becomes much less arbitrary if I go back many games and factor in results of each game, because ELO by its nature is self-correcting. So if I make a guess that's originally too high, if they lose games then their rating would fall quite quickly. Yes but the point I'm making is, there is no REAL mathematical way of estimating skill like this. Any numbers you come up with or what a win or loss does to this arbitrary number is still just that, arbitrary. As a player plays more and more games in a team setting, elo becomes more accurate supposedly, but with a sample size like TL Mafia, there is absolutely no way a skill rating like elo would come close to being an accurate representation. There are WAY too many mitigating factors such as the hosts, style of game (mini/normal/themed/whatever), roles available, who is playing in said game, etc. Theoretically this would matter less if people could play hundreds of games but that doesn't happen here.
Yeah. No one is saying this is correct at all.
|
On April 12 2013 01:49 Ace wrote:Show nested quote +On April 11 2013 22:40 marvellosity wrote:On April 11 2013 22:37 wherebugsgo wrote:
you don't know what I'm trying to say probably because you've come to a results-based conclusion.
You're incapable of seeing an alternate situation because you're basing your conclusion on the results of the game. Surprising that, isn't it. My belief is that a team I view as pretty one-sided would destroy the other team, the result supports my beliefs... yeah. Seems pretty reasonable to me. This game IS the alternate situation, and this game shows how the alternate situation pans out. But you're ignoring what actually happened. If myself or any other Townie hammers WoS on Day 1, or we let sciberbia get lynched Day 1 - both of which were close to happening this entire argument is meaningless. Looking at the end result without acknowledging how we got there and just declaring the entire scenario imbalanced is crazy  . Also saying 3 newbie scum vs a team with 3 "vets" is a bit dishonest - in this game myself, you and Palmar didn't even have as much impact as 3 or 4 other Townies. Imagine if we knew the teams were supposed to be balanced - after you flip green, then I get mislynched there is no reading of Oat's and Turnkeg's filters. Scibebria never makes his hypothesis post about the Axle wagon. No one even needs to bother re-reading the thread - lynch Palmar because of balance. That isn't Mafia play.
The WoS thing was quite interesting if someone actually hammered him. WoS was on L-2 when i asked marv "I think WoS is scum, should i vote for him now?". It was a silent message for everyone: "I find WoS suspicious, BUT FFS NOBODY DO NOT HAMMER YET!".
I think marv picked that up. I had no intention to hammer WoS and had no intention to actually lynch him at that point (although he was suspicious to me), i wanted to see if he does something stupid (as scum), or if scum does something stupid like hammer him (if WoS is town).
|
ITT: people who don't understand Elo
|
I think the only thing I didn't get about the hammer was this: (as it was my first Instant Majority Lynch game)
On April 11 2013 09:03 WaveofShadow wrote:Show nested quote +On April 11 2013 09:00 Ace wrote:
np. Lynching me was inevitable if going by sciberbia's theory (which turned out correct) and neither Oats or Tinkeg flipped. I just wished people read my scenario where if I was Scum why didn't I hammer you when I had the chance. Thought that was a dead giveaway but prp tunneled me to death and that got swept under the rug. Yeah my issue was as I said prp's thought processes were more or less exactly what I was thinking about you at the time; imo real hard to do something like that as scum. I figured you not hammering me would be because you'd be worried about how it would look bad rather than just go balls out for the mislynch but I guess not hammering (as can be seen in the scum QT) is just considered bad scum play.
|
United Kingdom36161 Posts
On April 12 2013 02:04 WaveofShadow wrote:Show nested quote +On April 12 2013 02:00 marvellosity wrote:On April 12 2013 01:58 syllogism wrote:
While you are doing arbitrary math, can you factor in the advantage town gets from knowing that teams aren't RNGed Well the thing is, it becomes much less arbitrary if I go back many games and factor in results of each game, because ELO by its nature is self-correcting. So if I make a guess that's originally too high, if they lose games then their rating would fall quite quickly. Yes but the point I'm making is, there is no REAL mathematical way of estimating skill like this. Any numbers you come up with or what a win or loss does to this arbitrary number is still just that, arbitrary. As a player plays more and more games in a team setting, elo becomes more accurate supposedly, but with a sample size like TL Mafia, there is absolutely no way a skill rating like elo would come close to being an accurate representation. There are WAY too many mitigating factors such as the hosts, style of game (mini/normal/themed/whatever), roles available, who is playing in said game, etc. Theoretically this would matter less if people could play hundreds of games but that doesn't happen here.
Well, 'who is playing in said game' is something that would indeed be factored in
It would be more a project of curiosity rather than something that absolutely establishes skill, obviously.
|
On April 12 2013 02:05 Ace wrote:Show nested quote +On April 12 2013 00:18 DarthPunk wrote:On April 12 2013 00:10 marvellosity wrote:On April 12 2013 00:07 DarthPunk wrote:
To be fair. Ace/marv/palmar all being town wasn't even a major reason that town won this game. Vets often play badly just like everyone else. So if we are talking about balancing 'vets' I really don't see the point.
Just as a general point, it's more the idea that ace/marv/palmar can organise/direct a mafia team well, and can exert significant thread influence for the benefit of mafia. So could a whole bunch of non vets in this game. I still don't see the point. You know something I have noticed though is that 'town leaders' from newbie games really quiet down in the transition to normals, I think a lot of that has to do with all the vet nonsense and thus they are less likely to contribute or take charge as they have in newbies. I quite like the idea of different players taking charge, or at least having the chance to take charge of a scum team.Having 'good scum players' in every scum team would become stagnant and deprives newer players that opportunity to grow and shine. Even if they don't succeed the opportunity is important. Case in point. Look at oats in this game fucking take charge in the scum QT and improve DRASTICALLY from his last scum game. He rose to the occasion and even if he lost that still has a heap of value. Furthermore, Who doesn't love a good comeback? or the tale of an underdog overcoming the odds and succeeding? I think there are a lot of intangibles to consider that are more important than some easy wins every now and then. agree with the bolded 10000000%. It's the main reason I take a backseat in many of the games I play. Let someone else develop and have a chance to take control of the game asTown or Scum. Some of my early game posts were just about keeping focus and avoidong what I thought were obvious mislynches. You guys did the heavy lifting.
I think the main problem with this is that they are not allowed to take charge. :/
|
On April 12 2013 01:49 Ace wrote:Show nested quote +On April 11 2013 22:40 marvellosity wrote:On April 11 2013 22:37 wherebugsgo wrote:
you don't know what I'm trying to say probably because you've come to a results-based conclusion.
You're incapable of seeing an alternate situation because you're basing your conclusion on the results of the game. Surprising that, isn't it. My belief is that a team I view as pretty one-sided would destroy the other team, the result supports my beliefs... yeah. Seems pretty reasonable to me. This game IS the alternate situation, and this game shows how the alternate situation pans out. But you're ignoring what actually happened. If myself or any other Townie hammers WoS on Day 1, or we let sciberbia get lynched Day 1 - both of which were close to happening this entire argument is meaningless. Looking at the end result without acknowledging how we got there and just declaring the entire scenario imbalanced is crazy . Also saying 3 newbie scum vs a team with 3 "vets" is a bit dishonest - in this game myself, you and Palmar didn't even have as much impact as 3 or 4 other Townies. Imagine if we knew the teams were supposed to be balanced - after you flip green, then I get mislynched there is no reading of Oat's and Turnkeg's filters. Scibebria never makes his hypothesis post about the Axle wagon. No one even needs to bother re-reading the thread - lynch Palmar because of balance. That isn't Mafia play.
I agree with everything said in this post. You cannot even begin to try to analyse mafia in a way that tries to account for player skill.
One important thing I think people are forgetting is that the game by definition doesn't allow players to play to their potential. A good townie on the right track gets shot and the rest of town proceeds to ignore everything he said, doesn't mean the townie deserves to lose ELO right?
now the answer is that for enough games you'd hit the true ELO anyway, but because of the complexity and inherent randomness of mafia you can never play even close to the amount of games needed to evaluate someone's true skill level.
So we make to with what we have. We use arbitrary metrics and judgement to decide player skill. But since skill varies massively even from one game to another for the same player, I honestly think balancing games is a bit silly.
|
On April 12 2013 02:08 WaveofShadow wrote:I think the only thing I didn't get about the hammer was this: (as it was my first Instant Majority Lynch game) Show nested quote +On April 11 2013 09:03 WaveofShadow wrote:On April 11 2013 09:00 Ace wrote:
np. Lynching me was inevitable if going by sciberbia's theory (which turned out correct) and neither Oats or Tinkeg flipped. I just wished people read my scenario where if I was Scum why didn't I hammer you when I had the chance. Thought that was a dead giveaway but prp tunneled me to death and that got swept under the rug. Yeah my issue was as I said prp's thought processes were more or less exactly what I was thinking about you at the time; imo real hard to do something like that as scum. I figured you not hammering me would be because you'd be worried about how it would look bad rather than just go balls out for the mislynch but I guess not hammering (as can be seen in the scum QT) is just considered bad scum play.
I'm rarely worried about how a hammer vote looks, but not hammering in this game as Scum probably was a bad play. Little time left in the day, a couple of wagons pushed, and tons of finger pointing probably means no one worth their salt is going to blame you for the hammer.
|
On April 12 2013 02:20 Ace wrote:Show nested quote +On April 12 2013 02:08 WaveofShadow wrote:I think the only thing I didn't get about the hammer was this: (as it was my first Instant Majority Lynch game) On April 11 2013 09:03 WaveofShadow wrote:On April 11 2013 09:00 Ace wrote:
np. Lynching me was inevitable if going by sciberbia's theory (which turned out correct) and neither Oats or Tinkeg flipped. I just wished people read my scenario where if I was Scum why didn't I hammer you when I had the chance. Thought that was a dead giveaway but prp tunneled me to death and that got swept under the rug. Yeah my issue was as I said prp's thought processes were more or less exactly what I was thinking about you at the time; imo real hard to do something like that as scum. I figured you not hammering me would be because you'd be worried about how it would look bad rather than just go balls out for the mislynch but I guess not hammering (as can be seen in the scum QT) is just considered bad scum play. I'm rarely worried about how a hammer vote looks, but not hammering in this game as Scum probably was a bad play. Little time left in the day, a couple of wagons pushed, and tons of finger pointing probably means no one worth their salt is going to blame you for the hammer.
really unfortunate timing and push choices screwed us there.
|
On April 12 2013 01:58 syllogism wrote:
While you are doing arbitrary math, can you factor in the advantage town gets from knowing that teams aren't RNGed
If you can tell how the players can guess, how the host groups and scores the players.
How the players, know what scope the host had to rebalance unbalanced teams by tweaks to the setup/roles which can vary from game to game.
How the players deal with the fact that whenever they do that and get it right regularly, then the host does it less, until they give up.
I would suggest hosts ought not say whats the maximum probability to reject a team.
I think you would find if you actually examined every possible 5,6,7 scoring team in the algorithm I suggested, that fairly often even with two flipped players, there is still very little actual information about who the third would be.
I am not sure, I could also try and factor in the advantage town gets when some scum players give up when they see their team mates. Well not so much give up but don't push as much other real life commitments out of the way to play as they would have if they had seriously viable team. Knowing that whenever you roll one of those hard games.. you were quite unlucky but that this is your chance.. to either shine if your the noob, become star by dragging a bunch of misfits over the line if you are the better, but not yet stellar player.
The math BTW was not arbitrary...
Also please note, I think i may have at this time a _personal_ preference for playing in RNG games. I will have to wait until I get ROFL stomped a statistically interesting number of times to be sure. Also at the moment I learn so much each time I play, winning/losing is meh. I do know however from having hosted RL games that were social games, (not mafia, but still social), what it took to keep my players happy. Reducing ROFL stomp frequency smells like it. I also hope that if I become better that I also wont get any satisfaction from beating an easy scum team.
(edit: not mafia == not any epic guessing game / social == played by people who largely use the game rules as an excuse, which game not specified as that would be misleading as most people dont play that game that way. )
|
On April 12 2013 02:00 WaveofShadow wrote:Show nested quote +On April 12 2013 01:55 marvellosity wrote:
Just as an arbitrary example for this game, if we take 1500 ELO to be 'starting' ELO with 2500 being totally exceptional and 500 being very poor, you might see something like the following (apologise in advance for over/under estimating players, more for shits and giggles)
1. iamperfection; town ELO 2000 mafia ELO 1000
2. WaveofShadow; town ELO 1500 mafia ELO 1500
3. Ace; town ELO 2100 mafia ELO 2500
4. prplhz; town ELO 1600 mafia ELO 1200
5. DarthPunk; town ELO 2000 mafia ELO 2000
6. Oatsmaster; town ELO 1400 mafia ELO 1300
7. Tunkeg; town ELO 1500 mafia ELO 1500
8. Palmar; town ELO 2400 mafia ELO 2000
9. marvellosity; town ELO 2100 mafia ELO 2400
10. Axlegreaser; town ELO 1300 mafia ELO 1300
11. raynpelinkeet; town ELO 1600 mafia ELO 1600
12. sciberbia; town ELO 2100 mafia ELO 1700
Average strength of mafia team: 4100/3 = 1366.6
Average strength of town team: 17400/9 = 1933.3
Best theoretical mafia team: 6900/3 = 2300
Worst theoretical town team: 15000/9 = 1666.6
Average town ELO: 21600/12 = 1800
Average mafia ELO: 20000/12 = 1666.6 I'm not sure what the point of doing something like this is with no real mathematical way of estimating skill.
Its a schtick for me to play with....
it will take little while though.. its 3am. Gdnight.
|
United Kingdom36161 Posts
btw, i'm not saying at all the ELO would be 'accurate', but it would be kinda interesting to see how the scores measured up to how the playerbase perceives different players. And then also to see how good the rating system was at predicting results of games...
like pretty sure Palmar's town ELO would be way lower than how he is perceived; he lost YANMM and LVIII as town, and as mafia I've handed him 4 other losses besides that in the last year (LIV, Death Note, NMM3, Hero), so his ELO would be relatively low for someone of his stature.
|
talking about Elo like this is more evidence that your entire mentality is results-based.
winrate is not everything in a game like this, because there are so many factors that go into a game. You can't compare a win in one game to a win in another; it's the same reason people fuck up meta all the time. Games are not strictly comparable one to one.
There are plenty of worse players who get carried and plenty of better players who get brought down. Yes, winrates may correlate with skill, but not nearly as strongly in other team games, because the sample sizes are so small. Average games played is probably less than 10. Good luck getting anything statistically significant out of samples that small.
(for the record; I have 7 scum games. You think you're gonna get anything useful out of those from an objective standpoint?)
|
Are we seriously trying to apply ELO to mafia?
|
|
|
|
|
|
EPT
