We propose a measure that quantifies the effort needed to compute a given stochastic average, considered in the context of molecular modeling. This "difficulty index" is defined in terms of CPU time, observed uncertainty, and a characteristic scale for the quantity being computed. This metric provides a focus for optimization and comparison of methods and, if applied broadly, could aid understanding of the impact of models, algorithms, implementations, and platforms on performance of molecular modeling techniques.