Yes, I optimized insofar as 1000 gave a better result than 500 and 2000, and looked better in the code. This strategy is not for trading, it was just the simplest that I found. For really trading it, it had to be professionally optimized and walk forward tested, and some better exit had to be programmed than the simple stop.
Yes, but EMAs and SMAs have a very different filter response curve than a standard lowpass filter. They also lag. A 100 bar SMA has a 50 bar lag. The lag of a lowpass filter depends on its steepness, but is usually far less than the lag of a moving average.
what is a "standard" low pass filter? what kind of response curve does it have? are you sure there are no windowing artifacts resulting in a sinc like response? are you using an FIR filter?
A standard lowpass filter has a frequency response that looks like a horizontal line that bends down at the cutoff frequency. You can see an example here: http://en.wikipedia.org/wiki/Lowpass In comparison, an EMA has a frequency response that first goes down sharply, without the horizontal part, and then tapers out. I don't have website with such a curve, though. I'm not using a FIR but a 2 pole lowpass filter, and indeed have some problems with artifacts, but it's good enough for filtering prices.
okay, it sounds like you know what you are doing, but I am not fully convinced that a first order RC filter (aka EMA) is so much worse than a second order one. But it could be true.
You optimized some trivial system and you got very good results. Nothing new about that. One problem is that this type of non-linear models are very sensitive to initial conditions. They are chaotic in a sense. This is the main reason they do not work in practice. Specifically, if you start from a nice peak and you get a good downtrend you have a cushion to work with. If you do not get the trend but choppy action for a while, you accumulate losses without a cushion. Thus, timing and initial conditions plus the fact that optimum parameter values change constantly turn these systems into big losers 99% of the time. Every trader with little experience knows not to waste time looking at such systems.
I like how your platform is built on Gamestudio. It was great fun back in 1996 making little games No idea it was still around.
Number of trades per day isn't as important as your usual target range. Unless you're only making a couple of ticks per trade, the notion of commission and slippage ruining your pnl is hogwash... Lets also not forget that slippage is usually ONE TICK on high volume contracts. That one tick of course happens on the buy side.