I am simulating a system to calculate mean of the reliability measure of my system. So, I am performing $n$ experiments where in each experiment I am simulating the whole system. Then I am taking an average of this $n$ results. Now I do this with $n=10,100,1000,\dots$. Each time I also calculate coefficient of variation ($CV$).
I notice that $n=10$ gives a lesser $CV$ than $n=100$. What is the significance of this result? Does that mean that it is not converging? How many experiments do I need to run to get a result which is almost equal to actual value.