Monte Carlo results represent an average of contributions from many histories sampled during the run. A statistical error (or uncertainty) is associated with the result. This number is of course very important but it cannot be taken into account alone. It is also very important to know how this uncertainty evolves with the number of histories. Indeed, this behaviour can reflect whether the result is statistically well behaved; if this is not the case, the uncertainty will not reflect the true confidence interval of the result which thus could be completely erroneous. MCNP is certainly one of the best codes that provide very detailed and efficient methods to deter­mine the quality of the confidence interval, as well as methods to improve the precision. Here, we just want to give an insight as an introduction of all the methods that can be used.