This paper talks about improving the sampling estimators in simulation results and catching rare events. Quite interesting discussions of Gaussian samples and approximation validation are also included. In most typical simulation works, modelers deal with approximation in terms of error bars and averages over 20 to 50 batch runs. The authors suggest that the methods they discuss can introduce smaller variance and thus increase the accuracy of the probabilities used.

The paper will be interesting to mathematically oriented readers who want to know how the different formulas can be condensed to answer certain questions. However, being more interested in the application of the theory described, I was disappointed that the authors only wrote a small section at the end about this; the accompanying algorithms are definitely useful, though. In fact, the paper is quite good in the way it explains the mathematics.

I would have liked to see more examples using the method (maybe a case study or two), as well as some situations where it is not useful. Although it seems to me that there must be computational overhead, that topic is not mentioned. Finally, the paper lacks a proper summary of the findings.