@Vector: I think it'd be better to have two groups of 5 role-claim to each other. Breaking into pairs just lets the SK WIFOM everyone into lynching the partner of whomever was NK'd.
Limiting the numbers keeps the SK from hitting the best targets. We can avoid WIFOM by just not taking everyone for their word as far as what the kills mean. Scum can only feed as much wine as the town is willing to drink.
This.
Also, my current plan (once other people get over here to discuss it...) is to create cells so that we cultivate disagreement/don't have two people who are likely to just go along with each other. My current model was each pair contains an analyst/active player + a tunneller/lurker.
The other thing is that I, at the very least, have a very hard time processing the amount of data flying around in a Mafia game. Splitting the town into two sets hardly reduces the amount of data... but 5 does something completely different entirely. We have 5 "informed minorities," 4 of which are guaranteed town.
Also, we want to squeeze the Webadict into as small of a space as possible. Right now we're working with 1-cells, and considering a 10-cell. It's easy to be a Webadict in either of those places.
In a 2-cell, with constant pressure from the townie partner and 4 other townie 2-cells looking at how the pair works together.... no, that's not easy at all. Consolidation of data for each player means more scrutiny on the Webadict, no matter who he is.
Essentially, because we've got only one scum, we're constructing scum-teams. It sounds ridiculous and batty, but my hope is that it will work.
Proposed list of pairings up in a little while. This job is more social than analytic, though, so I'll request help with altering pairs to generate the correct amount of friction when they go up.