Analysis of simulation results
Propagation of distributions - Determination of uncertainties by Monte Carlo simulation

The number of iterations must be sufficient to limit errors due to the stabilization of the numerical result. However, the number of iterations should not be excessive, as this will considerably increase computation time.

However, the problem of choosing the number of runs is secondary to concerns such as process analysis, quantification of input parameters and propagation of distributions. Indeed, it's better to have an uncertainty about the result linked to the number of draws made than to have an error linked to forgetting or unconsciously generating a systematic error.

Supplement 1 recommends one million prints. This recommendation should be adhered to whenever possible, but is not binding....