Mr. Subliminal: I think the problem is one that decent software programmers know full well. You are showing us a set of completed lines generated by a psuedo random number generator. Then you ask if we would trade them. In actual practice, one does not get to see the lines in advance and then ask that question. All one "sees" is one data point after another, and from that incomplete set of data, you have to decide whether or not it is "tradeable". This is the reason why I characterize the previous week of price data. It is not for the fun of it, but to determine whether or not price has exhibited startionarity during that previous time period. I make the assumption (right or wrong) that the percentages favor continuation rather than change (knowing that at some point, change is inevitable). I don't presume to teach anyone anything, but instead to point a finger (index) in a direction, hoping that enterprising people will become interested in looking "over there" in order to obtain an edge. If I had some way to see the completed lines as you have published I would certainly trade the hell out of them. Best Regards, Lefty
Let me reiterate that at this stage I don't wish to make any inferences regarding the market from this academic exercise. Nevertheless, I believe it to be both educational and interesting. On the contrary, if you read Sherry's definition of stationarity (below) carefully, you will see that my generated time series fulfills the condition of stationarity. "A stationary time series is one in which the underlying rules that generate the time series do not change over time." (The Mathematics of Technical Analysis p.9)
You have the algorithm - shouldn't that be sufficient? What do all the trend capturing tools try to do?
OK fair enough. First, I would have to backtest the algorithm. Lets say we start at .55 and move forward .2 per test. Naturally I would like to reserve the right to choose one or more values that tend to generate what I call trend, or even periodic trending behavior. Given the appropriate test procedure, I would be in a position to choose from a range of values that should generate (some) trending and thus tradeable lines (which I would hope to "trade the hell out of"). Regards, Lefty
OK, enough of the suspense. I spent many hours coding and testing Kaufman's Adaptive Moving Average (KAMA) on : (a) randomly generated "trends" and (b) "ordinary brownian" randomly generated prices. Result : It was possible to exploit the "trend" in (a). Not allowing for slippage and commissions, it showed very healthy profits. And this was without any tweaking, just using the default parameters suggested by Kaufman in his book. Using tweaks developed by Bruce (bdixon619) resulted in even higher returns. As expected, (b) yielded 0% return, tweaking notwithstanding.
So far people are seem to be saying that markets are not random. Can we all agree that markets are not "random" or do others have other things to say???