What's wrong with percentage?

[Kagami]

ps: balance is probably a bad thing to talk about when talking about randomness anyway. one thing i've been thinking about is how to code a program that would calculate win rates based on any role you put into the game, but in the end, what's more important than balance is swing factor and games with randomness are bound to have a higher swing factor (as opposed to, say, low power distributed among many members of each faction giving a game low swing factor, with mountainous setups having the lowest swing)

imo high swing factor is worse than an unbalanced game when looking at an individual game - i'd rather be on the 40 side of a 40-60 game than on any side in a 50-50 game that can potentially end N1

Muffin is like some kind of amazing sage of truthiness.

On a side note, I'd too thought quite a bit about writing a program to determine balance, but the problem is that the balance of a closed game is wildly influenced by the priors of the players, which is not something that can be nicely handled.

For not-too-large open setups, there are few enough states that a dynamic programming solution should work just fine. You can assume a [decide lynch -> claim -> lynch or redecide -> night -> loop] should capture the gameflow and define a state as the combination of all possible claims and investigation result claims for all living players.

[Kagami]

dynamic programming in the reinforcement learning context.

[AlwaysInnocent]

LicketyQuickety, I will address some of your statements first. I need to think a bit more about your other statements.

You would have to convince me that everyone plays the same roles the same way which would necessitate a "correct" way to play every role.

No, I would not even attempt to do such thing, because that is not what I believe. The reason I assumed it was to demonstrate the random chance of winning in situations that are otherwise equal.

While that looks appealing [to have a set strategy for any given role] such a thing is unattainable and I'd argue having roles that are dynamic rather than static can greatly increase a favorable means to an end of players actions to be done diversely as opposed to rehearsely.

It is because people play different strategies that the game becomes dynamic. Plus, as I said before, dynamic =/= random.

1.) I stand by my statement
2.) And yet the cycle where games are laughably unbalanced perpetuates.
3.) Like I said previously, with chance in roles, it makes NA more strait forward.
4.) Obviously, but why exactly is volatility a bad thing? Think of it in terms of risk reward.
5.) I'd like you to explain the word 'probabalistically'.

2.) Yes, this is very unfortunate, but this is entirely due to the fact that most mods refuse to think about their setups too deeply. Probably because they care more about theme games and creating wacky roles than a balanced game. They also tend to conform to one another, but that fails if the previous setup was flawed, too. It is something that we can overcome, if we want to. But it requires effort.
4.) Not necessarily, but volatility as a result of randomness seems worse than volatility introduced by subtle cause-and-effects.
5.) Now that I think about it, the word was probably misused. What I meant was something occurring purely by chance (i.e., something that can be determined by a coin flip or a die), rather than something that is caused by something directly and can thus be traced back (i.e., cannot be determined by a coin flip or a die, because then it would be impossible to trace it back).

[AlwaysInnocent]

Another thing I have not said about problems with introducing unnecessary chance is that you give more power to the mods to influence the game. It is impossible to verify that the mod is influencing the game that way or not. In an ideal game, mods have no power at all. This is possibly a controversial statement here, but I stand by it.

[chamber]

dynamic programming in the reinforcement learning context.

I've thought about this before and I'm not sure dynamic programming would be a very good fit. There is a lot of noise. The number of iterations you'd have to run for each generation would probably need to be in the tens of thousands. And it would still be reasonably prone to developing strategies for a specific set of roles(which works for opens). But you'd likely need to be willing to run it for days for each new open that you try.

[LicketyQuickety]

My whole push is that roles should be a fluid thing

yes i am aware that you are making assertions with no basis in reality

Heh... reality is subjective.

No

It's fucking not. If you can't make intelligible conversation, then just stop talking. Reality is not subjective, by very definition. This pisses me off that people think that "Nothing is real" is something deep that will make them seems smart. Anyone who has taken any epistemology or philosophy course knows that this is a ridiculous statement that does not hold up to any scrutiny.

What do you have to say about the perception of reality?