ADX Has it's Limitations
(I'm referring to Welles Wilders Average Directional Index in case you are a "newbie".) After many years of extolling the virtues of the ADX in articles and lectures all over the world I have become closely associated with this indicator. That's fine with me and I don't mind being considered the resident expert on ADX. It is an excellent measure of trendiness and a good indicator to be linked with.
However, I think it is a mistake to try and over work or become too dependent on any one indicator. If you were going to build a house you would need more than one tool and you wouldn't try to do it with just a hammer. The same is true of building systems. The ADX can be a very valuable tool if used correctly but it has some major shortcomings that everyone should be aware of: We all know that the ADX is slow. This is because of all the smoothing in the formula. The basic ingredients are smoothed and then the results are smoothed again. For example I think it takes more than 30 bars of data to calculate a 14 bar ADX. This smoothing makes the ADX slow but there is an even greater problem than just the speed of the indicator. The logic of measuring directional movement makes the ADX very reliable at certain times and very unreliable at other times.
A rising ADX is a reliable indication of a trend when there has been an extended sideways period before the trend gets started. Before all the high tech computer mumbo jumbo we used to simply refer to this sideways period as a "basing pattern". The ADX is most effective when it begins to rise from a low level (low = 15 or less). This low level on the ADX indicates that there has been a basing pattern for a while. This interpretation is contradictory to those users of the ADX who want to see the ADX cross above a specified threshold (usually 20 or 25) to indicate that a trend is underway. This technique would make the ADX even slower and means you would be confirming a trend and entering your trade long after the basing pattern was broken. But even if you were late due to your method of interpreting the ADX, following the ADX after a base pattern is still quite reliable. The potential problem I want to bring to your attention in this article is the action of the ADX after major peaks and valleys.
The logic of the ADX is best visualized as measuring directional movement over a moving window of data on a bar chart. If we have sideways data in the window followed by recent trending data (lets think of rising prices but it could be the reverse), the rising prices would show directional movement relative to the sideways data at the beginning of our window. The ADX would promptly rise and call our attention to the fact that there is now a direction in prices that should continue for a while.
However, if the prices rise for an extended period and then begin to fall sharply (a typical scenario) we now have a window of data that shows rising prices followed immediately by falling prices. The ADX formula measures the rising prices in the window and compares them with the declining prices in the window. Because the two trends are about equal they cancel each other and the ADX does not detect any net directional movement. The ADX now begins to decline indicating that it is finding no net directional movement in the period measured by the window.
As the window moves forward, eventually the older rising price data falls outside the back of the window so that the window now contains only the more recent downward price movement. The ADX suddenly begins to rise rapidly because the data window at this point contains only one trend. The problem with this new signal is that the downward trend in prices has been underway for quite some time and only now has the ADX finally begun to rise. This is obviously not a good point to be entering a trade to the short side. We are probably nearer the end of the trend than the beginning.
Remember that the ADX works best after a basing period and is
unreliable after a "V" bottom or top.
That"s all for now. I'll continue this discussion of the ADX in our next bulletin which should be out very soon.