And so cpo, there is the answer to your question. If you want to optimize a moving average, you must first realize that you don't need a moving average, and if you had one, there wouldn't be anything you could do with it anyway, and so the optimal setting would be zero.
Hello, again... Please refer to my "Brass Tacks" entry in this same thread for a background on my take on the movements of the markets. As for optimizing a moving average (MA), what you want to do is to determine the dominant cycle length of the market, then set the MA length (in code, of course, since the dominant cycle length is constantly changing) to exactly half of the dominant cycle length. Why 1/2? Well, look at the market as a continuously varying sine wave, then imagine if the sine wave was stuck at a length of exactly 14 days. That means that over 14 days, the sine wave travels from the bottom of one wave to its apex, and back to the bottom. With a MA set at, say, 14 days, you'd be measuring both the part of the sine wave that is going up (before it reaches its apex), as well as the part that is going back down. This makes the MA inefficient. But, if you set your MA length to 7 (in this case only...it'll constantly change as the market dominant cycle length changes), then you'll have a perfect MA that is showing only the increasing part of the sine wave as it's increasing, and only the decreasing part as it's decreasing. Perfect efficiency. How do you calculate the dominant market cycle length? You use a Hilbert transform (check out John Ehler's book, "Rocket Science For Traders" for a description) equation. If you want to get really fancy, (and can do some complicated coding), you can then take a MA of the first MA, figure out the lag between the two MA's, and add the length of the lag to the first MA. Viola! A perfectly synchronized MA with no lag! Again, check out the above mentioned book to figure out how to do this. Another trick you might want to try is to use a triple MA (three lines), each with a different length. The first would be the length of the dominant market cycle length, the second would be exactly 1/2 the length of the dominant market cycle, and the third would be 1/4 the dominant market cycle length. When the shorter MA crosses over one other MA, you're alerted to a possible trend, when it crosses over both of the other MA's, and the middle MA crosses over the longer MA (so you've got the longer MA on top, the middle length one in the middle, and the shorter one beneath), you've got a definite trend down. Vice versa for an uptrend. Of course, if your using a Weighted MA (WMA), then your length would be approximately 1/3 the dominant market cycle length. I wouldn't use an Exponential MA (EMA), simply because it's less responsive than a WMA, and tends to filter out too much of the trend you're trying to identify. That all being said, I must tell you that significant studies have been done that show that for larger markets (those with high volume), MA trend analysis is UNPROFITABLE for all MA lengths. The reason for this is that with a higher volume (more traders), there is more noise. In other words, traders are pulling their money out and putting their money into the markets for reasons other than the dominant trend. For instance, as more and more people began trading the DOW, you had people putting money in simply because they'd gotten a large inheritance, and wanted to do something with it. Or people who pulled their money out simply because they wanted to buy a car, boat, what-have-you. This introduces a large amount of noise into the market that overwhelms simple MA trading systems. Sometimes, having neophyte investors in the arena with you doesn't seem like such a great idea...notwithstanding the liquidity (i.e.: profits) they provide. Then, you also have to take into consideration those markets that are, for all intents and purposes, manipulated. Crude oil would be a perfect example...the price is not set by supply and demand, as most people would like to believe. OPEC, a cartel of oil exporting nations, gets together to decide the price (i.e.: price fixing on an international scale). Thus, trends are harder to pick up in markets like these, notwithholding the seasonal variations in price (which can easily be overridded by government action (dipping into stores to prevent untoward price increases), or by OPEC action (increasing or decreasing output to get the price they want), or even by non-OPEC nations (jacking up their output to take advantage of short-term price increases, which kills the trend).
MAs are a total waste to time, much easier and better methods of getting on board nice trends... simplier is better in this case!
Which is why I'd rather just look at the presence (or absence) of higher highs. Or lower lows. If they're there, I'm in an uptrend. Or downtrend. If they're not, I'm not. I've always been happy with Sperandeo's definitions of trend, trend change, and trend reversal. And they don't require moving averages at all. Simple. I like simple. But maybe that's just me . . . --Db
good stuff dgbro, and you have to wonder, if ma becomes less and less profitable as mkt becomes larger and larger, doesn't that really mean that mkt trends less and less?
so following that to it's logical conclusion, you would say that the long term trend for the market is choppiness.
Then it would simply be a matter of averaging the sucess rate of trendfollowing systems and overbought oversold systems and plotting them on a chart. When the overbought oversold average crosses above the trend following average you go overbought oversold. And when the overbought oversold average crosses back below the trend following average you go trend follower. And if you like to draw trendlines, you could just plot it all on the chart and draw trendlines so even though the market is not trending, you would still have a trendline to draw and you could still argue that you don't need moving averages and so everybody would still be happpy.
trend guy, Easier and better methods are a complete waste of time. There are much easier and better trends which don't require a method,