My attempts at creating an intelligent, adaptive system for the past 2 days have been met by frustration and annoyance. The idea is great, but trying to 1) computerize it and 2) get it to work together with my existing systems is a whole different story. Gaaahh.
One of the main reasons for it not working as well as I’d hoped is the adaptive indicator (the indicator that tells me if the market is trending). The adaptive indicator is a short term one, whereas the majority of my trend-following systems are medium to long term (to avoid getting whipsawed). Having a long term adaptive indicator is counter-productive, since by the time it indicates that a trend is present, the trend is already heavily under way.
Also, another problem lies with the complexity: having an adaptive system invariably results in many conditions in programming it. For example, I have to indicate rules like “if a trend is present (based on my adaptive indicator), then implement System X and Stop A. If not, then implement System Y and Stop B.” The multitude of rules makes system testing extremely slow, such that optimizing to find a good set of parameters is no longer possible. (well, I guess it would be possible, but testing would probably run for 4 to 6 hours – time I cannot afford) There is also the possible danger of curve-fitting.
In fact, the exercise got so frustrating today that I had to lie down on the grass on College Green to try to clear my head and get some inspiration. I came up with a methodology for doing my research: determining which periods were trending and non-trending, develop trending and countertrend systems separately, ensuring that they maximize ratios during their respective periods, then attempt to combine them. The last stage was where things fell apart. I’m going to have to spend more time on this.
However, if this whole adaptive system thing doesn’t work out, something good came out of it: I managed to come up with an adaptive stop-loss, which seems pretty useful from my cursory observation of it. When the markets are trending, it allows the markets to move in the anticipated direction, giving it lots of room. However, once my indicator tells me that the market has entered a sideways phase, the stops move closer to the price and take me out of the position for a profit.
As explained, the system enters long and the trailing stops are placed very far away from the actual price, but are always moving in my favor. However, once the trend is over, the stops take me right out.
A thought came into my head today as I was thinking about how frustrating this research was: Every failure I encounter now is just one more step towards preventing another costly trade in the future. I have to keep pressing on.