To what extent do optimal strategies exist?

jeep · Post Post #9 (isolation #0) » Thu Apr 01, 2004 1:30 pm

There is an optimal strategy, but because no one plays optimally yet, you can usually pick a strategy that has a better win expectation...

I will make my mafia simulator public after I deal with the board issues. It's really quite sophisticated... but it assumes perfect play, i.e. you cannot determine from the posts what people are scum/doc/etc.

I should make JEEPBot to play in games... So many projects I'd like to do.

-JEEP

jeep · Post Post #16 (isolation #1) » Wed Apr 14, 2004 11:17 am

mathcam: what is said MUST matter.

When bandwagoned, a doc will usually reveal. Now the town has to decide:
1) do they believe him
2) is it worth the risk of lynching him anyway?

So the fact that someone revealed must be relevant. Just like cop coming out. My program assumes that anti-leptons will be followed (which may skew my results if that is not always the correct play) It does branch to find out probability of town win if:
1) lynch and find cop
2) lynch and find not cop
3) do not lynch and find results accurate
4) do not lynch and anti-lepton's
It then decides if lynching is the correct play.

-JEEP

jeep · Post Post #19 (isolation #2) » Thu Apr 15, 2004 7:28 am

The goal of mafia is to play and have fun while trying to win. The optimal strategy is probably to never SAY anything. Simply vote and reveal if you are cop (or mafia posing as cop, etc). It's not a very fun way to play. I think it's important to study THAT version of the game to help determine what a balanced game is. I won't play THAT version of the game, though.

-JEEP