The nonzero gain coefficients of Sobol's sequences are always powers of two

When a plain Monte Carlo estimate on n samples has variance σ2/n, then scrambled digital nets attain a variance that is o(1/n) as n→∞. For finite n and an adversarially selected integrand, the variance of a scrambled (t,m,s)-net can be at most Γσ2/n for a maximal gain coefficient Γ