Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[TolyK]

Sure. I'll send some stuff over. It's in C# in VS2015.

[TolyK]

Alright, small update. I've gotten enough done to stimulate the simplest mafia game (n town, m mafia players, stimulate using random strategies), mostly due to having more time now. :p

[notquitethere]

Sweet. If you'd like any help or anything just drop me an email.

[TolyK]

So, some interesting preliminary results.
Keep in mind that there may be bugs in my program, however stepping through the code didn't reveal anything that wasn't fixed immediately...
Also, no reliable wifi on computer, so excuse the crap screenshot.


So, I ran a completely vanilla simulation of N townies and M mafia. Strategies were random: Townies voted random other people, mafiosi voted random townies and one killed a random townie at night. Ties in voting resulted in no lynches.
The results:

Spoiler (click to show/hide)

Horizontal axis: # mafia
Vertical axis: # town
Plot: % of games won by Town (actually, faction out of 100 trials).
http://m.imgur.com/gallery/tFUT75H
Yes, "screenshot"...

So, this came off as surprising, however it seems to make sense in that random strategies work poorly when ties give no lynches. Feel free to question any results. :p

[TheDarkStar]

I find it interesting that the current BM setup (7 town, 2 mafia) is listed as having a 0% chance of victory. On the other hand, it shows that even people new to mafia are better than a random number generator at finding mafia because the town does win BMs sometimes.

[flabort]

You'd better put some description on the imgur page. Your commenters think it's purely random numbers being generated, not a representation of anything.

[TolyK]

I don't really care about imgur folks downvoting or whatnot, though I really should give better descriptions.

However, what's next is much more interesting, I think.

http://imgur.com/a/SloeB

I should probably test two more voting methods (specifially, "revote until only 1 candidate" and "allocate votes from non-tied candidates", where I'd need to add special cases), but this seems to give the general idea that always lynching is much better than never lynching on a tie.


Edit: Updated with an even better method - exactly 1 of the candidates is (coin-flip) voted off. This is probably the most interesting/clear one.