I've held off posting this because IMHO the math isn't right ... what do you all think?
In particular, in their simulations they've varied the mean and variance of simulated lineups as if they're independent. In actuality:
1. Players with high variances also tend to have high means. There are few high-variance / low-mean players because they are lousy players and don't play except when there are no better ones available.
2. Player fantasy points aren't Gaussian, and the lineups are too small for the central limit theorem to kick in and reduce the impact of the lineup's higher moments.
So I take with a generous grain of salt the simulation result showing large profits for high-variance lineups with the "population" mean.