This is also known as : Mallow's C_p-criterion. In this model, the utility function is given by:

    C_p - n = (m - n_t) * Var(e) + (n - m) * Var(e_t)

    where

    Var(e_t) = variance of excess return of manager over benchmark, using all indices
    Var(e) = variance of excess return of manager over benchmark, using current subset
    n = number of indices in current subset
    n_t = total number of indices
    m = number of returns

Mallow's criterion calls for minimizing the utility function, i.e., StyleADVISOR chooses the subset of indices where C_p - n assumes its minimal value.