White Flag (Open Setup): 8 scum wins, 2 town wins, EV 47.8%, observed win rate 20%
Lovers Mafia: 11 scum wins, 6 town wins, EV 60%, observed win rate 35.3%
Grey Flag Nightless: 5 scum wins, 0 town wins, EV 45.9%, observed win rate 0%
I've been considering the long string of scum wins in Grey Flag Nightless to be something of an anomaly – it's clearly reasonable that it could happen by chance – but when we combine all the "one scum left = town win" setups that have been played often enough to give reasonable statistics, there's a clear pattern here. OK, so White Flag itself is a long-duration setup with a regular scum nightkill, and scum doing substantially better than EV there is possibly not too surprising as a result. But Lovers Mafia and Grey Flag Nightless are very limited in scum nightkills, something that in practice normally gives a huge advantage to town, and yet scum is doing considerably better than EV.
It's possible that all this is a coincidence, but I've now seen enough to suspect that there's something of a pattern here. It'd be useful from the setup-balancing point of view to figure out what's going on here, in order to help make future setups more balanced.
Some theories I have:
- The White Flag mechanic is typically used to allow more scum to fit into a vanilla game than normal, and with no power roles to centralise discussion, the large scumteams have a lot of control over the lynch vote, making it hard for town to get in early lynches on scum;
- The White Flag mechanic reduces the number of scum flips observed in a typical game (regardless of who wins), and towns are generally particularly reliant on scum flips;
- The White Flag mechanic discourages bussing-to-kill, helping scum in one of the two following ways:
- Bussing-to-kill is typically a bad strategy but is nonetheless commonly attempted, and discouraging scum from it therefore causes them to play better; OR
- Bussing-to-kill is a component in a WIFOM matrix that town have got used to allowing for, and thus a reduction in scum's WIFOM opportunities isn't hurting them as much as mathematics would predict.
Of course, this might all be completely wrong! Does anyone have their own ideas about this phenomenon? Can anyone think of ways to test any of the above theories (e.g. a setup that would interact with some theories but not others)?