Improving the efficiency of quantum Monte Carlo (QMC) to make possible the study of large molecules poses a great challenge. Evaluating the efficiency of Monte Carlo sampling, however, is at a rudimentary level and in need of new algorithms. Instead of the autocorrelation time as an efficiency measure for Monte Carlo simulations, we propose a direct method to characterize the movement of electrons in atoms or molecules during variational Monte Carlo computations. Further, the approach makes possible an efficient diagnostic tool to understand objectively many interesting issues in QMC. The usefulness of the method is demonstrated by comparisons among improved Metropolis algorithms and the original Metropolis algorithm. We also present an optimization method for choosing step sizes for Monte Carlo walkers. These step sizes are governed by the acceptance ratio of the electrons closest to the heaviest nucleus. Step sizes obtained for Ne and Ar are consistent with those obtained by the autocorrelation approach. Our study shows no evidence to support distinctions of core and valence electrons during simulations, and confirms that, in most cases, moving electrons individually is more efficient than moving ah the electrons at once. We find that "trapped" or "stale" configurations are due to a large quantum force, and a solution to this problem is suggested