Are you speaking of curvefitting or optimization in the positive sense of modeling market action accurately or in the negative sense of fallaciously modeling a random price sequence and regarding it as deterministic?
i get only little sense about them,and i want to make it clear that whether they are useful for the system trading or not,how to avoid their shortage......thanks
... always makes me a little leary. I like systems that show that they can work well within a range around the optimal parameter. Also, I would worry about stationarity - the results should be broken up over several non-overlapping time frames and analyzed. Just analyzing over one large block of time-series data, IMHO, can be very misleading. I wouldn't reduce a system to a single profit-factor number.