Ever since the new Mafia Stats Page was created, I have been following the numbers for my themes and some other themes I'm interested in, and I'm sure several other players have been doing that too. That page is even used to gauge balance for themes, but there lies a problem: What is the proper way to read those numbers to know if the theme is balanced or not? Let's begin with some examples (numbers picked at the time this post was made; let's not take game size in consideration for analysis' sake): Spoiler [/HIDE] So, which of those themes can be said to be balanced, according to the numbers? One would argue DN and Dinos are the most balanced, since all sides have a somewhat similar number of wins, which would mean Zelda and Kirby are unbalanced, since Village wins way more often as any other side. That thesis could be confirmed if you look at Caterpie and Oracle, where the main sides are quite similar in number of victories. But then, there's another way to interpret those numbers. Should a theme really be considered balanced if the win rate for each team is the same, regardless of any other factor? Some people could argue that Zelda and Kirby's village, although they have quite a gap in number of wins to the 2nd placed side, they doesn't actually have the upperhand when considered the overall wins. They win a lot, but actually lose more than they win. By that logic, Dino's herbivores and DN's Task Force have a really low win rate compared to their lose rate. This also can be confirmed when you look at Caterpie and Oracle, since village loses as often as they win. So, in your opinion, which one is the better way to gauge if the sides of a theme are balanced? All sides should have an equal win rate, or should village have a slighter higher rate due to having more players in and being the "good" side? What about solo roles? Let's discuss how you interpret the numbers from Mafia Stats, but please don't discuss if theme X is completely unbalanced because Y or Z, or theme W should be disabled because it's not balanced (leave those to the theme's thread). Let's focus on different ways to read those stats.
Good, I've wanted a topic like this. I made/balanced LOTR with the idea in mind that all sides should have an equal chance of winning. I see a ratio of 1-1-1, for example, as the "lower limit" of balancing while something like Myth, which has the village winning less than 50% of the time, would be the "upper limit" of balancing. Basically I think the win ratios should be between 1-1-1 and 2-1-1 where it's village-mafia1-mafia2. Lone roles are a bit difficult, if I look at Kronos (this is tricky, because the side Kronos has Uranus and Gaia in small games, so you have to look at medium size games where it's only Kronos), he wins about 45%-48% of the times he's in, assuming my math is correct. If that's true, I'd say that percentage for a lone role is pretty dangerous. A theme like myth (or even look at FF before Sephiroth spawns) has 4 mafia sides with the lone role, so I'd say winning between 16% and 25% would be good for lone roles. Rough calculations for now, they might be flawed.
Each side should have the same chance of winning, otherwise the village is usually going to win which by definition makes it unbalanced in the village's favour. Comparing the raw totals isn't a good way to do it though for obvious reasons.
The thing about village having a higher win rate is that mafia is often played as "all mafia vs. Village until village is out". Actually, I'd say it's rare to see games sith more than one mafia side not turning into "beat the village". For cases where this logic is applicable, we could say it's a 1 vs 1 situation, so village having the same win rate as any mafia could mean they are actually underpowered. @Hopkirk: I know raw totals are not the ideal way to check balance, but that's not the focus here, so I used them merely to build the examples in a way that makes it simple to discuss the idea. I used the raws, but that could very well be the numbers for a specific game size that someone wants to verify the balance.
I'd say to add a detection for when roles lists change, so that way stats are interpreted better and easier. Break down a total for each roles list... for example, in Kirby, put a combined total for games under 11 players. Then from 12 players to whatever number it turns to Roles 3.
I'd say break down and divide all sides by this: Number of games in which at least one role appears and wins the game / total number of games in which they appear This would help with themes like Elements which aren't exactly the most balanced (especially during small games). It seems like a stretch but it does seem like a manageable solution to the problem of not being able to generally compare without doing some manual work.
My intention was not to discuss how to break down the tables to analyze the balance. The idea is to discuss how to read and interpret the data after you already have your sample defined. For example, what does a 15-8-7 ratio to Village/Mafia1/Mafia2 at 15 players game would mean regarding balance? What are the possible ways to interpret those numbers, and what should be considered for each of those ways?
I would think Village should have the highest win ratio since they are normally the main faction. I actually think based on the data that omnivores are over powered. They should have more of a win rate like the carnivores. The other factions are of course going to win less often because they come at higher numbers and have many more obstacles in the way to win which is why they are such powerful roles. So in my opinion is should be village with the most wins and then equal with everyone else based on spawning. Rices 15-8-7 thing in the above post would make me think the theme was balanced almost to perfection.
Your forgetting something. Most of the windata isn't accurate because of the often afk wins and that different people play with different strategies. Wind Mage in Zelda don't win that very often because they often get haxed or get killed when they fakeclaim Link. No one really used to do that when Zelda was first launched and the claim Link as Vaati's Minion is very recent. so, because of the many variables, you can't really trust this windata. Anyway, it is natural for the village to lose more than they won. Village is the most powered side so it is natural that they win the most but as Rice said, the Village also loses. So it would be a total of all mafia wins vs Village wins.
Again, this thread is not about verifying how accurate the stats for theme X or Y are. The intention here is to discuss methods to analyze numbers after you already have your scope well defined and properly filtered. Also, your argument could very well be used for all sides. AFK people can make any side lose, and we can safely say it does, so if a side has more loses than it should because of AFK players, you can also expect it already won some games it shouldn't because of AFK opponents. Yet again I must ask that this thread focuses on how to analyze the numbers, not on how to correct them.
I aim for Village = All Mafias combined, or Village = Mafia1 = Mafia2 (but typically no more than a 3 way split) I find anything with a 40-55% split for village acceptable. They need some additional tools at their disposal because they are the uninformed majority. If they can't organize, they lose. If they organize, they win. If you make organization too easy (See old FE and old FF) then they take games without even trying. When I theme, I usually do Village, Mafia1, Mafia2, and some Solo+ roles, like Sushi Chef. Large games require that additional mafia faction, but there's not enough data on any game to see how it fairs. Breaking down Ocean for example: (00:08:18) *** OCEAN WIN DATA *** (00:08:18) Theme Played: 61 times (00:08:18) 1: Fish. Times Won: 28. Average Players per win: 11.21 (00:08:18) 2: Mammal Alliance. Times Won: 17. Average Players per win: 12.12 (00:08:18) 3: Shark Mafia. Times Won: 15. Average Players per win: 12.4 (00:08:18) 4: Cthulhu. Times Won: 1. Average Players per win: 13 The two mafia are pretty equal, except Mammals seem to be champs after 20 players, but overall they have their strengths and weaknesses. Village is taking just under 50% of games, while Mafia takes the rest, within my threshold for balance. Breaking down FE for example 00:01:29) *** FE WIN DATA *** (00:01:29) Theme Played: 240 times (00:01:29) 1: Heroes. Times Won: 91. Average Players per win: 12.49 (00:01:29) 2: Black Fang. Times Won: 49. Average Players per win: 14.69 (00:01:29) 3: Kingdom of Bern. Times Won: 41. Average Players per win: 16.85 (00:01:29) 4: Grado Empire. Times Won: 29. Average Players per win: 7.86 (00:01:29) 5: Begnion Empire. Times Won: 22. Average Players per win: 8.27 (00:01:29) 6: Kingdom of Daein. Times Won: 7. Average Players per win: 19.43 (00:01:29) 7: Tie. Times Won: 1. Average Players per win: 11 Based off that alone, Village is actually weaker than I like it to be, but that's perfectly fine. It's nice to see a theme where village actually has to work to win (And it makes so many people upset, which I thoroughly enjoy watching) Breaking down the stats farther, 88 games were played between 6 and 10 players Village took 37 Grado took 29 Begnion took 22 Ideally, in this situation, I'd like about 4-5 of those Village wins to be Begnion wins instead, but in that small of player count, a small change has a big impact. In games over 11 players this leaves 152 games: Village 54 Fang 49 Bern 41 Daeinn 7 and Tie 1 This is almost a similar split to the smaller games. A few of the village wins shifted to the weaker mafia and you've got a near perfect split between each major faction. Solo roles will ALWAYS have to be lower. If you give them an equal share, it ends up creating a super powerful, unstoppable, mega ultratron role that only loses if 42 players align in the formation of a rhombus in order to summon Galactus, the evil mega space robot that automatically kills the solo role. Despite the intent of this thread, I do think Stats should provide subtotals for each roles list