When browsing an old copy of STOCKS AND COMMODITIES recently, I had an "Aha!" moment, an "Of Course!" moment, a slap-myself-on-the-forehead moment. In "Building Automatic Trendlines," author Giorgos E. Siligardos makes the point that perhaps not all trends can be defined by trendlines.
Note: The charts on these pages do not feature up-to-date prices.
Annotated Three-Minute Chart of CMGI:
This problem arose when I studying the bounce off last summer's low for an article on the corrective fan theory. The first trending movement rose so sharply than it was difficult to draw the first trend, or fan line, off that rise. Was the trendline valid? I had questioned its validity in that article.
As Siligardos points out in a different example in his article, no peaks and troughs exist in CMGI's early morning drop on Friday, 3/30. How does one draw a descending trendline along the peaks when there are none?
Another problem exists. Siligardos also points out the subjectivity technical analysts sometimes employ when deciding what constitutes a peak or a trough.
Annotated 15-Minute Chart of the OEX:
There's more controversy than you might expect about the subject of the connection between trends and trendlines. For example, Clifford Pistolese might differ with my assessment that there was no uptrend during the illustrated period on the OEX. In USING TECHNICAL ANALYSIS, he describes how to determine an uptrend in prices, including all instances when the successive troughs are higher than the ones that preceded them. He goes on to add, "The locations of the intervening tops have no significance with respect to the establishment or confirmation of the uptrend." (5)
The "Chart School" Section of StockCharts.com specifies that a trendline is a "straight line that connects two or more price points and then extends into the future to act as a line of support or resistance." No requirement for corroborating higher highs in an uptrend or lower highs in a downtrend is made.
In MASTER THE MARKETS, Tom Williams asserts that, at the time the book was written, no documented scientific research verified that trendlines work. However, we all act on anecdotal evidence that they do, at least until broken. In that book, Williams cautioned that although traders might be able to draw trendlines mechanically, using the tools provided in their charting programs, they should not interpret them mechanically. Although even Williams seems to require only that successive peaks be lower in a downtrend or successive troughs be higher in an uptrend before constructing a trendline, he does appear to assert that some judgments be made about what constitutes an uptrend or downtrend.
Trendlines seem so easy, don't they? However, my attempt to draw a trendline (or fan line, for the corrective fan theory) under last summer's short, sharp first uptrend proved to me that they aren't so easy, even if I hadn't know that previously. I thought my problem lay in determining which trendline was the valid one, not whether there was one to be drawn at all. At the time I'd written the article on the corrective fan theory, I'd attempted to use RSI movements to corroborate or disprove the validity of a trendline I'd attempted to draw, a tactic that I've found helpful in other instances.
Siligardos would advise that traders classify trendlines as "potential" trendlines and "proper" ones. A proper trendline, in his classification, would require that both higher lows and at least one higher high be seen before a proper rising trendline can be drawn, and that both lower highs and at least one lower low be seen in a proper descending trendline. Otherwise, the trendline--and the trend it's attempting to classify--are only potential ones. The OEX trendline drawn above would have been a potential trendline and not a proper one under his classification. I can't quibble with that.
My "Aha!" moment had come when I had realized that perhaps the problem with finding the correct trendline hadn't been that I was placing it in the wrong place, but rather that that maybe one didn't exist on that chart. Uptrends in price can occur without a straight trendline construction being possible. Siligardos' article also reminded me to emphasize to subscribers that the fact that a trendline can be drawn doesn't verify that a price trend is in place.
My second "Aha!" moment occurred when I realized while scanning Siligardos' article that it may not be possible to draw a straight trendline defining what is clearly an uptrend, but perhaps that's because I'm working on an arithmetic chart. Aha! That was when it was time to slap myself on the forehead. Most charting services default to arithmetic charts rather than semi-logarithmic charts. On an arithmetic chart, the movements take on what are sometimes termed "parabolic" moves that curve sharply higher.
However, if such straight-up or straight-down trending movements, the so-called "parabolic" moves, are charted on a semi-logarithmic chart, straight trendlines often can be drawn! If such trendlines were then transferred to an arithmetic chart, they would show up as curves.
QCharts used to make it easy to switch from arithmetic to semi-log charts, but I haven't figured out a way to do it on my current charting program. I can't yet profit from my "Aha!" moment, but I hope that you can. When you see a move when prices are clearly trending, but you're unable to draw a trendline for the purposes of studying the movement because the move is too sharp for you to define peaks and troughs, try switching to a semi-log chart.
Conversely, if you've drawn a trendline along higher lows or lower highs, don't assume you've identified a trend in place. Make sure that you've also identified at least one higher high before you consider that rising trendline as defining an uptrend and at least one lower low before you consider a descending trendline as defining a downtrend.
If you're interested in reading further, Siligardos' article can be found in the
November 2006 issue of STOCKS & COMMODITIES. He includes code for setting up
trendlines that are defined not just by the parameters set up here but also by
specific percentage swings in prices from peaks and troughs, for those of you
who can input such code into your charting programs.