I think the best answer to this question is experiment: Try a simple system based on moving average crossing and you will see that optimization gives very different result for different securities and/or time period. Same with stochastic for example.
The truth is that every time series has its own features, and this same features change over time. The reason why volatility is more predictable than price is that it is not (directly) tradable, so you cannot use it to do arbitrage. If you look at the mathematical definition of technical indicators you will see that a low stochastic for example is a quite meaningless, since what it matters is the Fourier frequency of the price series (if any). If the stochastic is computed at the dominant frequency than it will give you periodic price up and down, but this frequency may vary overtime or not exist at all. Technical analysis is like to try to solve different problems in a math exam always with the same approach without even reading the text of the problem, just the title... Go on some finance job website like www.cityjobs.com and you will see banks and hedge funds do not look for any technical analyst but for time series and quantitative analyst to apply volatility models like garch and arch. Why would they miss such a great predictive tool such as technical analysis? Did you know big investments firms at wall street closed their technical analysis units and their former employees have now started their own consulting firms selling their secrets on the internet .. It is like all those websites in which people try to prove general relativity is wrong without even having an idea of what science is. They are full of posts and people arguing everyday about the next big truth missed by professional scientists. I think the only scientific true in technical analysis is the existence of support and resistance (static) and the correlation between volumes and prices Ly
