Monday, September 7, 2015

Dow Theory special issue: Why the Dow Theory uses daily bars, not weekly and for good reason

Primary and secondary trends for stocks, gold and silver and their ETFs miners unchanged.

Travis, one reader of this blog, recently posted a comment whereby he asked why I use daily bars instead of weekly bars to determine what allegedly is the long-term trend. What follows is my answer, which deserves a post of its own.

The primary bull (bear) market signals hinge on the properly gauging the lows (highs) of secondary reaction (not just “any” low, though significant it may look on the charts, as I have recently written here as many mistakenly believe). We also know that the time requirement for a secondary reaction to exist is at least 8-10 days (bars, when daily bars). However, since for a secondary reaction to exist we also need the extent requirement (at least a confirmed +/- 3% rally/decline), and the subsequent rally (pullback), most secondary reactions span a much longer time far exceeding 8-10 days.

The actual signal that follows the breakout of the secondary reaction highs (lows) may actually take several months. The last primary bear market signal of late August 2015 is a clear example of what I am meaning. In other words, the Dow Theory when working with daily bars is perfectly calibrated: Signals are sparse, and it is a far shot from following the short or even medium trend. It is wonderful to see how the Dow Theory by using “dailies," we actually determine the longer-term  trend. Furthermore, by using dailies and its associated “extent” requirement of 3%, we are able, while avoiding the short and medium trend, to exit at relatively “save” levels (that is near from the top) in most instances (at least this holds true for US Stocks, including the latest primary bear market signal). Of course, the non-parametric nature of the Dow Theory, plays a part in the usefulness of "dailies" in determining the long-term trend, since the absence of a parameter (i.e. we don't give a damn whether the lows of the secondary reaction to be broken are 10 days lows or 100 days lows, since we demand additional requirements), helps us get rid of non-qualifying lows. A simple example which, I hope, will clarify what I mean: it is not the same a breakout system (turtles-like) with just an n-days breakout requirement traded with daily bars, that the same system with an additional extent requirement which filters out many "n-days" breakouts. This is the issue that plagues "breakout" systems: in order to avoid too many signals, you have to either use "weekly" bars, or, with "dailies," you have to use a very high breakout number (i.e. 55 days breakout), which, nonetheless, continues to be very prone to whipsaws due to the absence of a filtering out of non-relevant highs (lows), which under the Dow Theory are filtered out thanks to the extent requirement, which serves to declare as the only valid lows (highs) those that qualify as secondary reaction lows (highs). Readers, now it's up to you to cogitate all this. For me, after years of mulling it over, this is as clear as day, but now you have to take your time until all this soaks in.

If we were dealing with weekly bars, then we should also adjust the minimum extent requirement. A weekly equals five trading days. According to Brownian motion (more here on Wikipedia: ), we should calculate the square root of 5 (5 days/ 1 day) in order to ascertain our minimum movement multiplier. In other words, if when working with “dailies” we require a minimum movement of 3%, when working with “weeklies” we should require 3 x 2.23 = 6.7% minimum movement for a movement to be meaningful. Furthermore, patterns are fractal, so the minimum time requirement would continue to be 8-10 bars (formerly, “days”). All in all, our new “secondary reactions” determined by weekly bars, would be too long term oriented. Our primary bull/bear market signals would be flashed too far away from the top/bottom. Thus, the 1987 crash would not have been avoided when using weekly bars, as a way to gauge the primary trend.

However, “weeklies” may be useful as a technical means to determine the secular trend. Let’s make a thought exercise. If when using “dailies” a Dow Theory primary bull/bear market signal is signaled ca. every 1-2 years, when using weeklies, we can expect signals to be signaled every 5-10 years, which is well attuned to the secular trend (multiplier of 5). I have some research (which will remain private for the time being), and the longer the time frame (by using, i.e. weeklies) the less noise, and hence the less signals. In other words, when being in a x 5 greater time frame, signals will not necessarily be signaled x 5 slower, but rather, you may expect even them to be signaled more than x 5 slower. Readers of this Dow Theory blog know that I am highly skeptical as to the usefulness for the average investor of the secular trend (more about his here and here). Having said this, if I had to rely on a method to gauge the secular trend, it would be technical (as with Dow Theory with “weeklies”), and certainly not, value/fundamentally based.

The use of weeklies as a way to establishing the secular trend may be useful for certain investors in order to avoid whipsaws when trading along the primary trend as determined by “dailies." In other words, one could wait for the secular trend to turn bullish (as determined by “weeklies”) in order to buy pursuant a primary bull market signal (as determined by the “dailies”). However, be advised, that by doing so you would have missed the monster primary bull market signal of mid-2009. If you wait to honor primary bull market signals (as determined by “dailies”) until the secular trend (as per “weeklies”) has turned bullish, you may avoid a few whipsaws, but you’ll miss a big piece of the action. 

By the way, if we gauge the current trend with weeklies, we have to conclude that the secular trend remains bullish. The current decline would qualify as a secondary reaction whose Friday, September 4th lows have not being violated yet.  Furthermore, no setup for a primary bear market signal has occurred yet, and hence, we cannot say that the final lows of the secondary reaction have been put yet. Nonetheless, I would not trust my fortunes (and timing) to a bullish secular reading, albeit such secular readings may be useful to some very long term oriented investors. Here you have a weekly chart of the Industrials, Transports and SPY (bottom), and the ongoing secondary reaction is highlighted by orange rectangles.

Dow Theory applied to weeklies: Just a secondary reaction against the secular bull market

Of course, the opposite holds true when applying the Dow Theory in shorter time frames. I also have some preliminary research that seems to prove that the Dow Theory can also be applied to lower time frame bars (i.e. 130 minutes bars). As with “weeklies," you have to adjust the minimum movement (which will be lower than 3%) in order to account for the lower time frame. Theoretically, signals should be given (when working with 130’ bars) 3 times for often than with “dailies” (as 130' x 3 = 390’=one trading day=1 daily bar), which would be ca. 1.5-3 signals per year (0.5-1 signal per year with dailies x 3). However, given that noise tends to increase with lower times frames, you should expect more signals, and hence more whipsaws, as the lower the time frame, the higher the noise and non-trendiness (Nassim Taleb, in his book “Dynamic Hedging," makes some interesting remarks concerning mean reversion, trend and time frames).


1) Daily bars are the ideal vehicle for gauging the long term trend under the Dow Theory, and to avoid crashes like the one of 1987 or to get aboard early as in mid-2009 and not miss a new bull market. 

2) Weekly bars are useful in order to gauge the secular trend (on a pure technical basis). However, they are not the “best” timing indicator, as it tends to be late at crucial market turnings.

3) The current “secular” trend remains bullish (a mere secondary reaction against the bullish secular trend) when gauged on a pure technical basis (with the Dow Theory), which may be of little consolation for those experiencing deep losses.

4) The Dow Theory seems to work at lower time frames (i.e. with 130' bars). However, noise increases and, while still profitable, key measures like profit factor, percentage win, etc. seem to deteriorate.  Further research should be conducted.


The Dow Theorist

No comments:

Post a Comment