Metabalance

[PolarBoy]

I thought people might like to hear some thoughts on the subject of game balancing, since there always is a bit of discussion about it here and there, and I had some thoughts recently that I wanted to share.

A common question people will ask of a game to determine it's quality is "Is it balanced?" It's a good question. But not a perfect one. The reason that balance is important is that one of the major points of a game is the tension that comes from the fact that you don't know who will win initially. It's a flawed question though, because balance isn't the important part of it. The important part is winnability. A player should be able to win, that way the tension of who will win is maintained.

I just lied. Winnability isn't important either. The only reason that winnability is important is because the player sees that he has a chance and continues struggling for victory. But if he didn't understand the game's situation completely, or something was hidden from him, the player might think he could win when really he had no chance.

It's stands then that what is important is not winnability itself, but the illusion of winnability that allows a player to believe in it.

This is how mafia is balanced. At the beginning of the game, the players do not know the entire setup of the game. They probably all have some information, but not enough to make any guess as to the likely outcome of the game. Bear in mind that a plyer could have no chance to win and still believe that he does. This is really based on a sort of balance anyway. We could call it metabalance.

Metabalance is basically this: While perhaps no instance is truly balanced, the direction of balance changes enough from game to game that there is no way to accurately predict a given instanfce, and therefore should be played as if one has a chance of winning.

There is a way to break metabalance. If every game was, ultimately, unbalanced in favor of one faction, players who found that they were not in that faction would ultimately decide not to play, as they would have an unreachable goal.

Actually this allows for a lot of intriguing things to happen in mafia games. For one thing roles that aren't in a game can still affect it. Cop variations are a good example. Because cop variations have been used in other games, a cop in a game could be a variation, so he can't be automatically trusted.

Anyway, those are my main thoughts on the subject.

[Stewie]

that was quite interesting, especially for people that are going to develop a game. I think I did it something along those lines

But sometimes, even roles that are virtually imposible to work with, like SK, give the player a small chance of winning. that small chance of winning is what allows the player to keep playing. Especially because if you actually win, you know you did something that is hard to acomplish.

[Antrax]

I think winnability, as you put it, IS important, because if I'm a player and I find out post-game I couldn't really win, it just seemed that way when I got my role, I'm going to personally boycot your games from now on. Some fairness has to be maintained, and that's best achieved by creating balanced games, in which every player's skill is what determines whether he wins or loses, and not the role distribution (though there will always be some luck in mafia, but the idea is clear)
Antrax

[MeMe]

I think winnability, as you put it, IS important, because if I'm a player and I find out post-game I couldn't really win, it just seemed that way when I got my role, I'm going to personally boycot your games from now on.

Absolutely. When writing a game, I ask myself, "If I were assigned this role, would it piss me off?" If my answer's "yes" (no matter if the "yes" would be immediate or at game conclusion) -- I leave it out.

[Dourgrim]

I have included roles that may have pissed the receiving players off before (see Asylum Mafia), but I'm not sure if that relates to the "winnability" of a game. *shrug*

[PolarBoy]

I made the case against winnability because the case for hardly needs to be spoken. Our natural sense of fair play demands it.

There are good reasons for real genuine winnabillity. For one thing it encourages better play. Whether one faction or another has a better shot it's important that the play decides the outcome. Otherwise a mod might as well randomize the roles as "you win" and "you lose".

My point was that a mod can only go so far with balance in the first place. I had no idea that the Serial Killer would win in toybox. I mainly put him there to speed up the game so that the mafia would be able to win in fewer days. You can't usually predict how a game is going to turn out. People will do things that surprise you all the time. My point was that it's no big deal, because Mafia makes up for a lack of balance with metabalance.

Basically, a player doesn't know where the imbalance will lie this time, so he should try as though he has a shot. Maybe he'll win. This swings hugely in the mod's favor.

Anyway, I'm sorry if I misrepresented the issue in the first post. I'm relatively new to all of this.

[Werebear]

I balance a game several times, (and remember I've only run two minis).

I consider the worst case scenario up to day 2. For mini 1, it would be either SK and mafia kill each other night 1, or SK kills Eddie the Computer (cop) and Mafia kills Gag Halfrunt (doc).

I consider role claims for the mafia. If I'm running one called "The 12 days of Christmas" and the players find out that days 10, 11,and 12 are mafia... mass role claims would ruin the game. In HHGG2, I specifically chose Max Quordlepleen as GF, because I thought someone could distance themselves from the other two mafia (the waiter and the dish of the day), especially with people claiming to be an investigator that didn't get results.

I examined each role individually to see if any was too powerful. For instance, I originally had the Dish of the Day able to suicide to take out two players of the mafia's choice. This was more powerful than I wanted, one dead mafia for two dead players of the mafia's choice (along with a normal kill, of course). I also originally had Marvin never get a good result - only depressed results. I changed the first one to the mafia had a chance of eliminating themselves (the mafia actually coudl have been over night 2), and changed the second one so it had the possibility of being an excellent role - if Marvin rolled a 10, he would have gotten the most complete investigation I could have assembled for him. If I could have added the player's real name and address, I would have. *grin*

last, I looked at the power of the roles en masse. In HHGG2, I had three mafia including a GF, the other two had extra kills (the waiter could poison, the Dish of the Day could suicide to take out two other players). This is a powerful mafia. Night 1 they could have taken out 3 players by suiciding, and the next night they could have taken out two if the Waiter wasn't lynched... so by day 2, the mafia would have killed 6 players. To balance that, I made the town powerful too... 1 sane investigator, 1 insane investigator, and three potential investigators, a two-shot doc and a one shot vigi. I still think the town was a wee bit more powerful than it should have been, but I also think the mafia would have won if the waiter had poisoned night 1, and hadn't been lynched day 1.

Even all this aside, it's still a matter of how balanced it feels. The best-balanced mafia could be over in 5 days, a poorly balanced one could last weeks and be full of intrigue and suspense, depending on choices made. You really just have to make roles, and work out how things might go in your mind.

[mathcam]

I definitely think every player should have a non-negligible shot at winning, but I think that's about as far as it goes in my book. Exactly in your terms, I think there's a natural metabalancing that the game designer doesn't even have control over. If you get a role that's hard to win with, then you've exceeded all expectations by wining with it. I'll take serial killer any time any one wants to hand it to me. I know that I'll have a hard time winning, but I know that if I win, it'll count for more.

How can you really measure someone's ability at mafia? Certainly not wins/losses. Certainly not how long you can last in any particular game. Any of these can be artifically lowered or decreased without actual talent in the game. The "perfect" measuring tool is the nebulous and very hard to define "How well has the person played with what (s)he was given?" The best we can calculate this is essentially something to the extent of "What is the relationship between the expected number of times this player should win compared to the actual number of times (s)he did win?" This leads to my example before. In this scoring system, you don't lose a lot for not winning with a serial killer, but you gain a lot if you do win. That's why I don't mind playing one.

I realize that not everyone thinks about the game (or games in general) this way, and I definitely agree that you should play to win every game, not to maximize your overall performance, but I don't think my above interpretation of how to play affects this at all.

Cam

[Antrax]

I don't see why everyone think the SK is so impossible...
Antrax

[mathcam]

Not at all impossible...just severely handicapped relative to the mafia. Almost always, whatever the mafia's chances are of winning a particular game, the Sk's chances are significantly smaller.

Cam

[shadyforce]

I was just thinking. Imagine we had 3/4 mods come up with a theme for a mini. Then we have that same mini played many times simultaniously each with different players. Players won't be able to see how other minis are unwravelling. Ten once they are all finished, we compare how each player with similar roles performed. We can see how accurately we can tell if a game is balanced or not, or whether there is a lot of luck or indeed skill involved. It would be a mammoth task to set up and carry out but it would be very interesting.

[mathcam]

There was a huge thread about this on the old board. I think Captain Blicero was thinking about running his Texas Mafia game again, but with all new players. I'd be really interested in seeing that happen. Cap'n? Thoughts?

Cam

[shadyforce]

Maybe a game that has been permanently deleted.

[PolarBoy]

The problem with SKs is that they're usually added almost as an afterthought. The game is usually written so that the conflict is centered around the Mafia and Town, and the SK is added mainly to mess things up for one of them.

The traditional SK role usually is pretty limited, having no one on his side and generally having more difficult win conditions than anyone else. In a mini, the town usually has to eliminate 2-4 players to win the game, the mafia has to eliminate 6-10, the Sk has to eliminate 9-10. It's pretty easy to see who the odds are in favor of.

Generally though it seems like a good idea to do something to help the SK out. One modification I like is allowing him to lie low from investigations while not killing.

[Macros]

those numbers are unfair and missleading though.
i still agree sk is right down there for winnability, but basing chances merely on who has to be eliminated..
eg - bar masons, the town only can trust themselves (sk in similar situation) and the sk can pull a lot more uber cool stunts and gambits than a mafia can (imo)
mafia know each other, vastly unblaancing those numbers.
sorr,y that was pointless, i'll leave this logical sensible discussion now.

[mathcam]

I often give serial killers an immunity to the first time their killed at night, and in Quantum Mafia, also provided them with a "safe" role claim...i.e. a role that fit other pro-town roles and was not taken by anyone else.

Maybe a game that has been permanently deleted.

Say, for example, CaptainBlicero's Texas mafia game?

Cam

[mith]

I love Serial Killer roles. For some reason, they almost always make it to the endgame in my games, and then die the night before they could have won.

I've never gotten to play a SK role that was anything like normal. I've had a Mad Scientist role twice, whose only purpose is to piss everyone off as far as I'm concerned (can't kill, just keep people from posting or something strange like that), and a ninja role which might have been cool except it was somewhat spoiled by having to be played over AIM to prevent nights from being too long, which limited my choices.

None of that has anything to do with this thread, though.

When I'm balancing a game, I shoot for about even on Mafia/Town chances, with the SK having probably around a 5-10% chance; I've thought of a few ways to make that a bit higher, but I haven't tried them out yet. If there is more than one Mafia, the Town's chances tend to be slightly higher than either of the Mafia groups. I do agree with the general opinion, the game should be winnable for everyone, even if it is only a small chance. There have been quite a few games which annoyed me because I found out at the end that I couldn't possibly have won.

[Sugar]

I was just thinking. Imagine we had 3/4 mods come up with a theme for a mini. Then we have that same mini played many times simultaniously each with different players.

I remember reading this discussion on the old forums... I'd really enjoy participating in such an experiment.

[MeMe]

This could be very embarrassing -- finding out that someone excelled in a role I crashed with -- but I'd love to participate in the same-game-played-simultaneously-by-different-players experiment, too.

[mathcam]

I love Serial Killer roles. For some reason, they almost alwas make it to the endgame in my games, and then die the night before they could have won.

Hmm, I think I

was

one of those in your Mafia 11, and then died the night before they could have wone.

And I agree that if we could make the logistics work, it would be really fun. I even think it would be fun if all the players could watch the games happening at the same time. Day one would be interesting..."Well that claim makes sense...in the other game, Bob claimed that same role....they wouldn't both be lying, would they?" "Sure he would. Bob claimed that role

because

it was claimed in the other game. He thought it would be a safe claim," etc.

Cam

[God]

that seems like a really cool experiment to try

seeing how the game plays out more than once with different players...

i'd definetly participate, but i think it'd be better if the players in each game couldn't see the other game. perhaps a secret passworded part of a board to try it on or something...

[CaptainBlicero]

Yeah, I'm going to be bringing Texas Mafia back, probably in a week and a half. I figure I'll throw it up in the Free Market Queue, since I'm only going to allow members who joined post-crash to play. After that, I'll publish a summary of the first game and compare it with the second. It should be very interesting.

As for metabalance, I think the idea of discretionary power is very important. By discretionary power, I mean being able to impact the game in a tangible way. This is why SK roles are so much fun: even if you know your chance of winning outright is pretty low, you can still take great pleasure in watching your body count pile up. So, I think if you create a role that has a very low chance of winning, you've got to give the role a high amount of discretionary power to make up for it. Because you definitely want to make sure that players don't feel screwed when the setup gets revealed at the end.

[Stewie]

I'm only going to allow members who joined post-crash to play..

I think I joined 5 days before we moved to this board, can I play?

Anyways, about the 3-4 games at the same time I think the idea is that you can't see the other games, but the only way I see to do this is too actually make a new board with, let's say invisionfree. But I rather play it here, because invisionfree crashes each 2 days or so...

[bloojay]

yay for post-crash joiner's like me

[PolarBoy]

Stewie, the reason only post-crashers are allowed is because pre-crashers would've had the opportunity to read the game thread. If you never did, then you're probably clear.

I guess I should stop bringing up roles from Texas Mafia in discussion threads for a while.