In some versions of Quill’s Simon Sez trading systems a pair (or a triplet) of moving averages is included. He favors 13/50. But I find that pair doesn’t backtest any better than many other fast-slow variants. In other versions of SS, he uses the widely used triplet of 50, 100, 200 to estimate favorable/unfavorable market conditions (as does Weinstein). But I think that makes for a very cluttered chart.
As everyone knows, moving averages (MAs) come in many versions: simple, weighted, exponential, double exponential, Hull, etc. They differ from one another according to the weighting of the components of the series used to calculate the average and the extent to which the average is smoothed. Almost any MA type can be forced to provide nearly the same signals as any other MA type simply by adjusted the lookback period over which it is calculated. Some lookback periods are favored over others. But unless rigorous backtesting is done, the choices are arbitrary. As with choosing which fishing lure to cast, choosing which MA to plot is a matter of preference and habit with the upside being that what one uses frequently is likely to be effective simply because it’s what one gets used to using.
My favorite MA type is the triangular MA, because I like the swoopy curves creating by giving more weight to the middle potion of the time-prices series from which the MA is calculated than to the most recent prices, as is done with weighed and exponential MAs. But as I said, the choice of MA type and lookback period are arbitrary unless one has done the backtesting to prove otherwise. Anecdotally, a 5-period Triangular MA gets me in and out of trades close enough to the bottom and to the top, most of the time, to never need changing out to something faster, slower, or better. I’ve expermented with using pairs of trianular MAs. But I dislike the clutter created by plotting a second line and prefer the signals created by price crosing over/under a single MA than the signals created when a faster MA crosses over/under a slower one. Again, this is yet another ‘Chef’s Choice’.
Now comes the part where things get complicated. If one is trading off of HA bars (Heikin Ashi) , one is never dealing with exact prices, but a smoothing of them. If one is applying MAs to a plot based on HA bars, then that introduces yet another level of abstraction and inexactness which can raise this question, “Did a ‘Buy’ or ‘Sell’ signal occur or not?” Here’s a chart that illustrates that problem.
Look at the dogi in the first third of the chart. If one is using ‘price crossing the MA’ as the signal, then no signal has occured, right? Same-same the following day. But now look at a chart done with hollow candles.
Huge diff, right?
Retrospectively, the smoothed version of prices, aka, HA bars, kept one in the trend IF one is only using crossovers for signals. The unsmoothed version, aka, hollow candles, had one getting out at the open on the 15th on a day of rising prices but missing out on four further days of gains. Given that one can’t trade yesterday’s market, a decision has to be made between the two chart formats and their implied rule sets. Dojis are important, and there are plenty of studies that confirm their usefulness and reliablity as trading signals. Therefore, I’d make this compmise.
Plot wirh HA bars (to retain their greater trend clarity) and use crosses of price over a 5-period triangular MA UNLESS a clear doji is printed. In which case, GET OUT at the next day’s open.
I haven’t done the backtesting needed to justfy that rule, nor do I intend to. From looking at lots and lots of charts, it simply is what makes sense for the Chicken Little trader I am to do.
Charlie
