I have done plenty of trading on optimised MA/MA crossovers etc in my early days. I find that there is no ideal MA. Optimise one MA on a given data sample and test it on different data sample, it will not be the best. On this data sample we will have different MA as optimised MA.