Starting from the premise that the increase in the number of values is made with values belonging to the same population, it is possible that this increase shows differences in the assumed or estimated distribution for the sample. For example, the increase in data could show that an initial assumption of linearity presents exponential characteristics.

In my opinion, according to the law of large numbers, the use of a larger sample size should converge to the real MSE value.

but How can I fix it? cuz the values of MSE should decrease while the sample size increase Pablo Constantino

Try using a suitable distribution such as normal or beta

The statement "the values of MSE should decrease while the sample size increase" is not necessarily true. It depends on the distributions used for the random variables and the mathematical model for the simulation. It is difficult to know with the information in the question.