Moving averages, Breakout strategies, etc. are no match for the Dow Theory
No new from
the markets. Primary and secondary trends unchanged. And hence, I will further
cogitate about the Dow Theory. Such moments of market lull are good to rethink
one’s premises.
The more I
know the entrails of the Dow Theory, the more fascinated I am about it.
Clearly, Charles Dow was a genius,
since mere chance cannot devise such a perfect way to determine the trend. And due credit should be given to Rhea and Schannep whose work has further streamlined Dow's insights.
When Dow
Theorists appraise secondary reactions (hereinafter, “secondaries”), we
routinely mention the “time” (at least 10 calendar days) and “extend”(at least
3% move) requirements. We are so used to it that we don’t give it a second
thought.
However, by
bringing together the time and extent element, the Dow Theory has nothing to do
with its two closest competitors: Moving averages and sheer momentum (be it
relative or absolute measured as a percentage change).
To begin
with, the extent and duration requirements used by the Dow Theory to appraise
secondaries are both non parametric. And this is a very good thing.
If we want to
determine the trend using a moving average, we are just using the duration
(time series) element, but have no regard to the underlying extent of the move
(hence false signals on week trends and multiple moving average crossovers). A
moving average is blind to the previous price structure (i.e. whether there was big
momentum before the crossover, weak momentum, how many days of big momentum,
how many days of vanishing momentum, etc.). Furthermore, a moving average is
the best example of a parametric system. Parametric systems, as I have explained here and here, are undependable and
highly unstable.
Sheer
momentum strategies have no regard to time
element and focus blindly on the “extent” element. Thus, a stock that may
have been languishing for 11 months, suddenly rockets in month 12, and suddenly
ranks among, i.e. top 1 percent. However, sheer magnitude without regard to the
time elements (what did the stock do in the 11 preceding months) can be
misleading and prone to false signals too. Furthermore, if we demand a
determined percentage change threshold (i.e. it must be up 30% in the last 12
months), we are again toying with parameters, which may backfire in real life.
As an aside, my personal take, is that the best way to use momentum is as
relative strength, that is, by just taking the top “x” percent of a given
universe, irrespective of their individual percentage gain. By doing this, we
try to get rid of parameters as much as possible, being the only “weak”
parameter the ranking percentage. In a future post of this Dow Theory blog I
will write about “strong” parameters (dangerous, as they are prone to
“overfitting” and share all the flaw of parametric systems) and “weak”
parameters (which, are less dangerous, and less prone to over fitting). To say
that I will select the top 10% high relative stocks of the S&P 500 is much
less dangerous than selecting the stocks that gained more than 26.58% in the
last 12 months.
Lack of
distinction between “safe” and “unsafe” parameters misleads many system
testers. For many years, I erroneously thought that any parameter was bringing my system closer and closer to curve
fitting. This is not so. Thus, even if we considered the 3% minimum extent
requirement of the Dow Theory as a “parameter”, we can see that it is a very
“safe” one. You can successfully apply the Dow Theory, if in instead of 3% you
take 3.5%, 4% or 5% as your definition of minimum movement...You just will get
less secondaries and fewer signals (and more ample initial stops) but the system
will perform roughly equally well. You will stand to lose a little bit more on
some (not all) losing trades, but you will be less whipsawed. This proves that
the 3% requirement is not a “parameter” or, if you deem it to be, a quite
harmless one. Mutatis mutandis we can
say the same of the “time” extend for a secondary to be appraised.
So both
momentum and moving averages have two salient flaws.
a) Tend to be
quite “parametric” (the worst offender are moving averages).
b) Each
ignores the complementary element which would confirm (time should confirm
extent, and extent should confirm time).
The Dow
theory is non parametric. The time requirement (i.e. 10 days for sec reaction)
and minimum extent (3%) are just this: minimum
requirements to weed out noise. Called them if you want “weak, harmless,
parameters”. However, there is no
preset level of what constitutes the time and extent element of a secondary
reaction which is the cornerstone of the Dow Theory (as primary bull
and bear market signals depend on the breakup or breakdown of the highs and
lows of secondaries and hence it is vital to properly ascertain them). You can
have secondaries with just 4% downward movement and 20 days formation period , or with just 10
days and a 7% decline, all combinations are possible, and hence we see the beauty of the
Dow Theory that it adapts to market conditions unlike inflexible moving
averages. As an aside, the Dow Theory stops (secondary reaction lows) are not set at arbitrary levels (i.e. 20 days lows) but rather correspond with market action. A 20 days lows stop can be too narrow if the market has been raging of late (or too wide, if the market underwent a huge rally), whereas the Dow Theory stop based on the lows of the secondary reaction "adapts" to recent market action. And hence it can be larger or narrower depending precisely on market action.
Therefore, the
net superiority of the Dow Theory is built in into the system: We depend on two
non parametric variables: time and
extent. So minimum time should confirm minimum extent and viceversa….So we can find
the principle of confirmation even inside the entrails of the Dow Theory.
If you add “external confirmation” (the classic rule “two indices” should confirm)
with the “internal” one (“time and extent” should confirm), then you have in
your hands, what I think is the most reliable tool to gauge the trend and keep
whipsaws to an absolute minimum.
All in all,
the Dow Theory:
a) Is non
parametric.
b) Relies on
“internal” confirmation: Time and Extent should confirm each other.
c) Relies on
“external” confirmation. Indices should confirm each other.
Now go to my
post analyzing the Dow Theory performance. It should not surprise you that ca.
70% of the trades are winners (something unknown in other trend following
systems) with a remarkable win to lose ratio. Something unknown in typical
trend following systems (we were taught that you either have a high batting
average with a poor win/lose ration, or vice versa, but you cannot have both;
well the Dow Theory proves this is simple wrong).
If you
further cogitate, you will start to see that the Dow Theory is also clearly
superior to traditional “breakout” systems (i.e., buy if today’s close is
higher than the highest high of the last 50 days). Do such breakout systems include an “extent” element? Do such systems contain a dynamic exit point adapted
to previous market action (as secondary reactions are)? Or they just have dummy
parametric set stoplosses? Although, this is the subject of a future post, I
can loudly say that I am not a turtle, as such systems are parametric
flawed, and disregard the “extent” element, and hence the scares they give to
their practitioners every now and then (and huge drawdowns which very few
investors have the stomach to digest)
Once this
post soaks in, you’ll have your “aha” moment.
Sincerely,
The Dow
Theorist
No comments:
Post a Comment