Friday, May 24, 2013

Dow Theory special issue: Dissecting the classical/Rhea Dow Theory record




 Part I


In previous posts of this Dow Theory blog, we occupied themselves with the evaluation of the Dow Theory performance, as explained here. Furthermore, we broke down such data into secular and bull markets as presented here and here.

One thread in common was that all these studies focused on a comparison of the Dow Theory versus buy and hold on year-end basis. Thus, if on December 31 of any given year the Dow Theory was up 4% and a position taken in the Industrials was up 3%, we would say that the Dow Theory outperformed buy and hold by 1%.

However, such analysis did not focus on the specific transactions taken in pursuance of the Dow Theory buy and sell signals (which correspond to primary bull and bear markets).

We want to know basic statistics concerning each Dow Theory signal, namely:

·        How many signals have been flashed in the last +115 years?


·      How long does the average position taken in pursuance of the Dow Theory last?


·        What percentages of signals end up winners? And losers?


·        What’s the average gain?


·        What’s the average gain in winning transactions?


·        What’s the average loss in losing transactions? 

     What's the win to lose ratio?


·        What percentage of time I am in the market?


·        What’s the profit factor?


·        Do secular bull and bear markets affect the Dow Theory signals?

Such statistics are vital to master in "real time" the application of the Dow Theory and, as far as I know, they have never been published before with this kind of breath. 

Accordingly, I plan to start a new series of posts in this Dow Theory blog dealing with all these issues.

This post will focus on the generalities of the Dow Theory signals and, thus, I will not distinguish between secular bull and bear markets. In further posts, I will narrow down my examination to secular bull and bear markets.

A final caveat: This analysis concerns what I label the “classical/Rhea” Dow Theory. In a nutshell: This is the Dow Theory as advocated by Dow Theorist Rhea in the ninety's thirties. It only uses two indices (the Industrials and Transports). The “classical/Rhea” interpretation of the Dow Theory tends to disregard values (versus the more secular oriented Schaeffer’s version) and focuses on the technical side of the market. You can find more about the classica/Rhea “flavor” here.

The “classical/Rhea” version of the Dow Theory stands half-way between the too long and secular oriented (in my opinion) Schaefer’s version and the slightly more reactive Schannep’s version (which avails itself of 3 indices). Personally, I tend to like more Schannep’s version, since it clearly outperforms the “classical/Rhea” version. However, the “classical/Rhea” version is not to be overlooked, as it has an excellent track record and has been the object of study by many brilliant minds like Russell of the Dow Theory Letters’ fame. Thus, I see Schannep as an improvement over something that what was close to perfect. Nevertheless, in order to understand Schannep we must be fully conversant with the “classical/Rhea” Dow Theory.

This Dow Theory blog, little by little, is completing an in-depth analysis of the “classical/Rhea” Dow Theory as our foundation stone. This is our starting point, and before we delve into Schannep’s intricacies, we must know inside out the “classical/Rhea” Dow Theory. This knowledge of the “classical/Rhea” Dow Theory is to be broken down into two areas:

a)     On the one hand, to master the technical rules (grasping the nuances of a secondary reaction, primary bull and bear markets, etc.), so that one can  spot the signals on the charts and determine current market conditions.

b)     On the other hand, to be fully acquainted with the empirical record. When faced with an ongoing position, the investor must be able to know how similar signals performed in the past. While the past is no guarantee for the future, a record spanning +115 years is not to be disregarded out of hand. 

Fortune punishes the unprepared. The more the investor knows the rules and the past, the better the odds for success.

The record I have used spans more than 115 years. In begins with a buy signal on July 12, 1897 and finishes with a sell signal on June 4, 2012. The last primary bull market buy signal was flashed on January 18, 2013 (details of such a signal here) and is not included in my analysis, since, as I write these lines, no sell signal has been flashed.

As no “official” record of the Dow Theory exists, I derive my record from these sources:

From 1897 to October 1956, as contained in the book “Technical Analysis of Stock Trends," 8th edition, by Edwards and Magee, and reproduced by Schannep in his book “Dow Theory for the 21st Century” (pages 26-29).

From October 1956 to date, the “classical/Rhea” Dow Theory record has been reconstructed by Schannep in pursuance of the classical rules and not his own.

The whole record can be found on Schannep’s website, thedowtheory.com


Well, after this lengthy introduction, let’s get started with the meat of this post. 

CLASSICAL DOW THEORY VITAL STATISTICS

Total number of transactions (trades) taken: 38

Comment:

We can see that 38 trades in +115 years has nothing to do with short-term trading. The positions taken in pursuance of the Dow Theory rules tend to last significant time. If we take into account that almost 1/3 of the time one is “in cash” standing on the sidelines, it is clear that no hectic trading is necessary. This implies a low turnover, and with it, lower transaction costs.

In my opinion, the Dow Theory strikes a perfect balance between being the guy holding the proverbial hot potato (and being killed by a +50% drawdown when a vicious bear market sets in), and being a stressed short term trader.

Average duration of each transaction: 712 days.

Comment:

I quote what I wrote in this post:


The average duration of each transaction taking according to the classical Dow Theory lasted 712 days. This is slightly less than 2 years.

The median duration amounts to 565 days, which is roughly 1.5 years.

The shortest investment lasted only 60 days (year 1990).

The longest investment lasted 2799 days (secular bull market 1900-1998).

All the instances lasting less than 180 days (ca. 6 months) occurred during secular bear markets (save one position, which lasted only 60 days and was taken in 1990 within the secular 1982-1999 bull market). Therefore, when assessing the probable duration of a new primary bull market, it is advisable to try to put the nascent trend in context. If it is born within a secular bull market, it is likely an “above average” duration. If, conversely, the market is under the spell of a secular bear market, the odds favor shorter trends. While I acknowledge than in real time not so easy to label the market a “secular bullish or bearish”, we can make our guesses.

Thus, we can conclude:

1.      The average duration of the positions taken according to the classical/Rhea Dow Theory last ca. 2 years on average. If we took the mean, we’d derive shorter time spans.

2.      While occasionally we may encounter trends (or better said, positions within a trend) that last 4 years or more, this is by no means the rule but rather the exception.

3.      Shorter trends (and accordingly positions) tend to occur during secular bear markets. In technical parlance, we would say that the market is prone to false breakouts.

 Winning trades (positions) versus losing trades

29 transactions resulted in profits versus 9 losers, that is 76.3% winners, versus 23.7% losers.

While this is not carved in stone, it gives us a pretty good indication of what to expect in the future.

A healthy percentage of “winners” is important. I know that in trading circles, it is said that the percentage of winners is immaterial, and what really counts is the win to lose ratio. If when winning one makes 10% but when losing one loses 40%, then 76.3% of winners can lead directly to the almshouse. Thus, many traders favor systems (normally breakout systems) with a low percentage of “winners” (i.e. 30%) but a high payoff ratio (i.e. 4 to 1). However, in my opinion, this is true (and to a very limited extent) for shorter term trading and when dealing with leveraged positions with narrow stops. Most trading systems place stops at quite arbitrary levels and most of such stops rather than being technically determined are dictated by the amount the trader is willing to lose in each trade.

Things are different for the Dow Theory investor. Our stops, as I have written here and here are clearly technically determined. Investors along the primary trend are not supposed to be leveraged. Thus, any attempt at leveraging Dow Theory positions and, with it, narrowing stops to “contain” risk, would be a mockery of the Dow Theory.

Such a high percentage of winning transactions (76.3%) tells us something important about the accuracy and solidity of the Dow Theory. In essence, the Dow Theory is a breakout system. One buys penetrations of previous highs and sells violations of previous lows. As I wrote above, breakout systems tend to have a low percentage of winning trades (30-35%), since many breakouts are false ones and prices immediately reverse and hit the stoploss. Thus, the Dow Theory clearly excels at determine which “highs” are the relevant ones to be broken (buy signal) and which “lows” are the relevant ones to be used as stoplosses, as not all “highs” and “lows” are relevant under Dow Theory. We can confidently say that the Dow Theory does an excellent job in separating the wheat from the chaff and in this fashion, it manages to achieve an astounding 76.3% of winning trades thanks to the combination of finding the “right” breakouts and placing the proper technically defined stoplosses.

Finally, we shouldn’t underestimate human nature and our psychological frailty. While a system with a 30% of winners can be profitable,  it takes a toll on the investor’s physique. I know this firsthand. A system with a 30% winning trades will result in long strains of losses, which eventually destroy the trader’s ability to stick with the system.

A healthy 76.3% reduces the odds for a long string of losses and, thus, it is “tradable” in real life; not just in trading books or trade labs.


Average gain in each position: 32.33%

Comment:

32.33% average gain is a respectable figure. We should bear in mind that this average gain is made in somewhat less than 2 years (average trade duration). If we also bear in mind that almost 1/3 of the time we are flat with no long position, 32.33% is an even more remarkable figure.

As with any trend following system, the Dow Theory does not catch the absolute tops and bottoms. Thus, we can confidently say that buy signals (primary bull market signals) are flashed with a delay of ca. 10% (that is 10% after the market lows). The same applies to sell signals (primary bear market signals). Thus, if the average gain stands at 32.33%, this means that the total average trend should be increased by 20% from top to bottom (10% lost from the top and 10% lost from the bottom). This implies that the average trend implies an average gain from bottom to top of ca. 52.33%. Alas, only 32.33% can be “extracted” as profits by the Dow Theorist.

While, this may seem outrageous for an outside investor, this is the price we have to pay to be reasonably sure that a new trend (or trend reversal) has been established. I don’t know of any timing or trend following method that can spot earlier trends without jeopardizing the overall profitability. In investing there is no free lunch.

32.33% is a respectable figure for the buildup of capital. If we assume transaction costs of ca. 1% (slippage +cost of buy and sell and custody if the shares are not held in the street’s name), such 1% is a modest amount of cost and doesn’t eat up much of the raw performance. 


The average gain in winning transactions stands at 54.12%

Comment:

Here we focus on the winners. The average gain in such instance stands at 54.12%.

This is an important figure in order to establish the risk-reward  ratio of each prospective position. Depending on the specific setup,  each primary bull market signal will have a different stoploss. When a primary bull market signal is flashed, our initial stoploss stands at the last recorded primary bear market lows. Such stoploss may be as tight as 4.2% (i.e. buy signal of 5/13/1980) or a loose as 20.3% (buy signal 10/7/1982). Of course, if one uses the Schannep’s stoploss, which I explained here, our stop should never exceed 16%.

In any instance, the “risk” in the reward risk ratio is very variable depending on the specific setup.  Once we know the denominator (“risk”), be it tight or loose, we can make our best guess as to the reward. If we are very conservative, we can take as reward, the average gain of 32.33%. However, technically it may be more appropriate to take as the likely reward the average winning trade, instead of the average trade. Why? Because, we only have two outcomes in a given trade: Either it is a winner or a loser.

If it is a loser, we know how much we stand to lose (our Dow Theory stop). Here we don’t talk of likely losses, but of hard-and-fast  figures.

If it is a winner, then….it is a winner. Do you get it? If it is a winner, our proper reference is the average gain made in winning positions, not just the average gain (which includes both winners and losers).

However, I err many times on the side of caution, and when assessing the reward risk ratio of a new Dow Theory position, I just take the average gain, instead of the average gain of winning trades.

Average loss in losing transactions: -6.10%

Comment:

What else can I say? It is self-explanatory. A loss of -6.1% clearly fits within the parameters I gave concerning the initial Dow Theory stop (at the primary bear market lows). I wrote above that it ranges from a modest 4.2% to a whopping 20.3%. Of course, this is the worst-case scenario, since, as soon as secondary reactions develop and the lows of the reaction are not broken, the Dow Theory stop is raised, and, accordingly, the amount to be lost diminishes (and in most instances, profits are locked in). Serious investors should have a seamless understanding of the Dow Theory trailing stop. The first step is to read this post and study its graph.


Averages can be misleading. Here you have the actual figures of all losing trades:

          Day/Month/Year   Price Indust   Pct Lost
SELL
13/05/1940
       137.5
           -3.56
SELL
26/04/1962
       678.68
           -3.96
SELL
25/02/1969
       899.8   
          -4.51
SELL
26/01/1970
      768.88
        -10.62
SELL
19/10/1978
      846.41
           -2.32
SELL
03/08/1990
    2809.65
           -4.28
SELL
25/06/2002
    9126.8
           -4.81
SELL
29/09/2008
  10365.45
        -19.33
SELL
04/06/2012
  12101.46
          -1.57






Avg Loss
    -6.11

Well, the maximum loss was suffered on September 29, 2008 (-19.33%).

We also observe that the Dow Theory avoided catastrophic losses in all instances. Even 2008s -19.33% loss compares favorable with the ca. -50% meltdown the market experienced. In all other instances, losses are few and well contained. Thus, Dow Theory acts as an excellent capital protector. The best offense (returns) is a good defense (contained losses). As I wrote here when evaluating year-end returns:

Investors get blinded by performance. However, in real life, the investor is killed by draw downs. A 15% average performance is worth nothing if, somewhere along the road, there is going to be a drawdown of -50%. Buy and hold is nice in theory, and it may work provided the investor has deep pockets (staying power) and psychological fortitude. However, in real life, very few investors possess both attributes at the same time. Thus, the publicized return figures of many investing strategies are not attainable in real life because the investor cannot endure the draw downs. If the average retiree needs to draw 4% off his capital annually (and this is a very realistic and even modest assumption), a drawdown of 50% in any given year, will force him to draw 8% if he wants to keep his expenditures intact. Of course, he can cut with expenses, but as we well know, this is not an easy feat. Even if the retiree manages to reduce expenditures by 25%, this implies a withdrawal of 6% while being in the midst of the draw down. As a result total equity would be reduced by 50%+6%, thereby remaining only 44% of his original capital. A draw down of such magnitude is akin to a black hole. It is very difficult to escape from it. In most instances, the retiree will finish by eating up all of his capital. Game over for him!

The astute reader should be asking: “What happened in 1929 and 1987, I don’t see such years in the spreadsheet?” Here is my answer: the Dow Theory managed to get investors out of stocks before the market crash. In other words, a sell signal (primary bear market) was flashed shortly before the crashes began in earnest. Here you have the details of the two positions which preceded both crashes:

             Day/Month/Year    Price Indust    Pctg gain
BUY
07/12/1923
      93.8

SELL
23/10/1929
    305.85
226.07

BUY
21/01/1985
   1261.37

SELL
15/10/1987
   2355.09
86.71

As you can see in both instances, the investor was forewarned on time by the Dow Theory. Furthermore, in both cases, the investor managed to lock in sizeable profits (226.07% in 1929 and 86.71% in 1987). You can find more details about the remarkable job the Dow Theory did in 1987 in my post “Revisiting the 1987 crash," which you can find here.


Well, this is enough for today. In a future post I will examine other statistics such as “win to lose ratio”, “percentage of time in the market”, “profit factor” and “secular bull and bear markets”.

In the meantime, have a nice weekend.

Sincerely,

The Dow Theorist

No comments:

Post a Comment