Awhile back i had a thread asking about everyone's toughts on moving averagea. Pretty much everyone said they were either worthless, or at best people said they might look to them to dtermine context, but their actual buy/sell determinations are based other indicators/triggers/price action. Take a look here: http://www.robertwcolby.com/RelativeAdvantage.html The two moving average systems were at the top of the list as far as effectiveness of everything tested. Thought on this article? Its a bit dated, and it might be dinged in other ways as well. Curious to hear all yalls perpectives!
Data stops at 2001. Markets change. And are these curve-fit? How did 3 day, or 6 day MAs work? Did performance drop off drastically? And how much was the advantage over BnH, especially with trading friction/taxes thrown in? You'd think with something that simple, if it was that great, hedgies and others would easily discover these systems and trade them into oblivion.
Without having read all the details from The Encyclopedia of Market Indicators that your link references, I'm guessing he simply backtested from 1900 to 2001 without using an out-of sample test period. That is to say, he may have just taken a whole bunch of market indicators, ran them on that 100 year period, and ranked them. IF that is the case (and I admit this is an "if" but I don't own the book so I can't look at the details more closely), then some of the indicators will outperform others simply by luck. It's the same concept as overfitting data when selecting parameters for a model, for example. Here he may have overfit the selection of indicators... especially if he tried many different time periods of moving averages (eg 3 day, 4 day, 5 day, .... etc and the 5 day just happened to be worth mentioning.) What would be more useful is if he selected a few top indicators using part of the data and then showed that they CONTINUED to outperform during the rest of the data (ie the out-of-sample or test data). If he did this, then I would be more interested. Do you own the book? Can you give us more details on the methodology?
Why SHOULDN'T MA's work: they are derived from market data. The question is how often do they fail? Everything works if done consistently, but not always in meaningful ways.
I use them and have for many years but not as signals themselves. 9ema, 30wma and 90wma across three time frames. There are no two identical trade plans and you have to discover what works for YOU and not what conforms to ET opinions.
Statistical, dont own the book sorry. Wish I did. Yea, you just cant glean enough from what's in the link to tell if the conclusions/results are anything meaningful. Just thought that they came out on top in anything was funny given their general perception. wish i know more.
If you read Colby's most recent remarks concerning his study of moving averages, he goes on to say that they basically don't work in today's markets the way they did when he originally published his findings. He blames HFT/Algo, etc. It's in the article you posted. Used the correct way and with other means of verification/filtering, MA's have a useful purpose. Remember: Price makes the average. The average does not make the price.
It is human nature to want the milk without the cow. The real question should be: "how do I use the MA (or insert any other TA study here) in order to correctly to take a trade entry". IMO, once you have an informed baseline of how to use an indicator correctly, it is much easier to tailor the study settings and time frames to suit your needs. John Bollinger once told me at a reception that 95% of the people using Bollinger Bands don't use them the way he designed them or intended. I tend to find a similar theme with new trader clients I take on.