The rarest, most massive dark matter halos are quite challenging to simulate numerically but also relatively easy to understand theoretically. I will describe how several aspects of massive halos may be understood using simple, self-similar collapse models. Combined with the (known) statistics of Gaussian random fields, this model makes predictions for the abundance and clustering of massive halos, with no adjustable parameters. This allows us to extend theoretical predictions for the halo mass function into regimes beyond the range of N-body simulations but accessible to ongoing cluster surveys.