The more I think through this stuff, the more I think it's going to boil down to PRs saving us even with optimal play by town and errors by scum. If we go into D7 with 1 or 0 townies alive, I think there's no way for town to win, even if our 1-shot-deathproof person is town. These eight situations are everything I think could happen going into D7 with 2 or more townies and the various D5 lynch/N5 kill patterns:
(3T, 2M, 1SK)
(3T, 1M, 1SK)
(3T, 0M, 1SK)
(2T, 2M, 1SK)
(2T, 1M, 1SK)
(2T, 0M, 1SK)
(3T, 2M, 0SK)
(2T, 2M, 0SK)
Maybe it's useful to agree on which of the possible combinations leads to even a potential win from here.
(3T, 2M, 1SK) -> Status quo (Happens only after no-lynch, with scum failing to kill, so not likely, and almost certainly worse than now).
(3T, 1M, 1SK) -> Possible win if we lynch scum D7.
(3T, 0M, 1SK) -> Possible win.
(2T, 2M, 1SK) -> Probable loss, depends on PR distribution.
(2T, 1M, 1SK) -> Probable loss, depends on PR distribution.
(2T, 0M, 1SK) -> Standard 3-person lylo, possible win.
(3T, 2M, 0SK) -> Standard 5-person lylo, possible win.
(2T, 2M, 0SK) -> Probable loss.
Does anyone think these evaluations are incorrect?