|
PureBytes Links
Trading Reference Links
|
I wrote this in reply to the subject question from a member at another
trading message board: are the behaviors of the markets random or not?
What you believe about the answer, impacts what automated, mechanical and/or
experiential/judgmental trading systems and tools you should choose for your
trading...if you trade or invest at all. But maybe none of that really
matters much?
Posted with permission of the person who emailed me the question.
- - -
IMHO the behaviors of the equity markets are not completely random ...even
though I think that much, but not all, of Taleb's "Fooled by Randomness"
book has merit, is thought provoking.
1. News impacts the markets. That effect is certainly not random,
although the news itself may be. The effects strangely precede the news,
often. There is a wide variety of types of news, obviously, so when you mix
the combinations up, the resulting impacts can appear to be random when in
fact they are a complex combination. News can be about an individual
security and impact only that one, or impact many others. News can be about
peace or war or disasters or economic statistics or earnings forecasts or
executive changes or even live CNBC interviews of security analysts and
company executives.
Sometimes the news has been so anticipated that its pre- and post-release
effects are small.
It will be very interesting to me to see what happens if we go to a full
blown war against Iraq (the war started on a small scale several months ago
via CIA and NSA paramilitary operations). Have the markets already largely
discounted this?
Infernally, events and therefore news can be total surprises. What happens
if Saddam is suddenly overthrown or assassinated? What happens if Saddam
attacks first? What happens when we capture or kill bin Laden? What happens
when Arafat dies from natural causes or is assassinated? What happens if a
major terrorist attack, of 9-11 proportions or larger, occurs anywhere in
the world?
No mechanical or automated trading system can anticipate the news nor
related pre- and post- news impact. A well designed system will get you out
of trades going rapidly against you, at worst, with stop loss triggered
market orders.
2. Manipulations. The market is constantly being manipulated, at least at
an individual security level. You could also say that the Fed's interest
rate actions, and the Fed's limited control over the value of the dollar
manipulate entire markets.
Manipulation is certainly not random. It occurs when one or more
institutions and/or large investors decide to accumulate or distribute a
significant number of shares or contracts of a particular security or group
of securities.
Rather than write a book here about this, recommend that you read some of
the 1870 - 1940 or so published books about trading. Many have explanations
of how pools operated. Today's mutual funds, pension funds, hedge funds,
large companies, large investors and traders, etc., operate in the same way,
artificially keeping a security's price high when selling, and artificially
keeping a security's price low when buying.
One system that could work quite well, with safeguards, would be one that
could detect institutional distribution or accumulation, and trade in the
appropriate direction. Problems include how deviously clever the software
is, and master traders are, that perform this for some institutions.
3. Sectors. The business outlook and to a lesser degree, the past
quarter's performance of leading companies in a given sector, influence the
prices of all securities in that sector.
Certainly not random.
4. General economic conditions, and related consumer (and thus, investor)
confidence. Certainly not random, and these certainly effect the markets.
5. Crowd psychology. The actions and reactions of various crowds of folks
in the markets. Related to all of above, plus to human psychology of crowd
behavior. Certainly not random. May be nearly impossible to predict,
though.
6. Extremes. The occurrence of statistical tails in market data is far too
frequent and far too large, to be random.
7. Fundamentals. Fundamentals do not always impact market behaviors,
especially for individual securities.
One reason is that fundamentals are mostly past history and may have little
to nothing to do with future performance.
Another, as we've seen with WorldCom, Enron and Global Crossing,
fundamentals can be manipulated. For those of you who understand
accounting, this manipulation can be quite legal yet deceptive.
Still another, by the time fundamentals reach public records, insiders have
already taken advantage of the good or bad surprises and the securities have
already moved. Martha Stewart is simply a person who allegedly got caught.
Perhaps many thousands never get caught.
And most of all, fundamentals change too slowly to be use to traders, and
are often too vague, too imprecise, for entries and exits.
That all said, sustained long term fundamentals do influence the long term
trend in value of securities. This behavior is certainly not random.
8. Randomness. The sheer random occurrence of particular traders and
investors and news and all of the rest of above, acting simultaneously at a
point in time on a particular security.
The end result of the combinations of all of the above can give the
appearance of a totally random market.
But that is simply because you would have to back out all of the nonrandom
effects, and then look at what is left.
- - -
Another way of explaining randomness and forecasting, a possible analogies
for trading?
Say you are a market research analyst for a manufacturer, wholesaler or
retailer, and your job is to forecast product sales so that inventories can
be at the right levels at the right times.
One form of forecasting, primarily used for this, is exponential smoothing.
This method results in a seasonal and trending forecast, in combination.
Plus some stuff left over called noise, that may or may not be random, but
that must be treated as if random. Noise is the critical determinant of
safety stock levels, because by definition it is the part that is not
forecastable. Excepting that its bandwidth is forecastable with some % of
certainty, which equates roughly to the inventory service level targets.
99% service level means carrying enough inventory to statistically be able
to fill orders, covering 99% of that random bandwidth (3 standard
deviations, if my memory is correct).
Or, you can forecast with one or more leading indicators, if the products
happen to have such. Car repair parts can be forecasted years in advance,
to some degree, by knowing new car sales now. Weather is a leading
indicator for air-conditioning, commercial refrigeration, heating repairs
and sales, etc. There is randomness in the leading indicator(s) that you
will have to either smooth or treat like is done in exponential smoothing
techniques. This form of forecasting is called, of course, regression
analysis or multiple regression analysis. One of the difficulties is
finding leading indicators that relate in a causal manner to that which is
being forecasted. Less difficult but very challenging, is discovering the
details of the relationships.
Or you might use both methods for different time periods and different
purposes.
Don't remember ever reading about the above approaches being tried on
securities, but am reasonably sure both methods have has been tried.
What this is meant to explain is, the behaviors of a market are a wild
combination of randomness and nonrandomness. If TA can help separate these
into seasonality, trend and noise in some fashions, then TA is helpful for
trading. The randomness becomes a bandwidth around the possibly
forecastable part.
9. Time period. The shorter the time period considered, the greater, I
suspect, is the randomness in the data. The longer the time period
considered, I suspect randomness becomes less important.
10. Dominance: randomness or order? If randomness is not dominant,
attempts at forecasting prices and volume are still specious. Less
randomness does not necessarily equate to accuracy of forecasts.
That’s because of the complexity of so many things interacting at once. And
because the things that trigger the start, end and extent of price and
volume changes are not particularly predicable. And because the extents of
the impacts of those on human crowd psychology are also not particularly
predictable.
11. Randomness is irrelevant in the end. Money and risk management are the
key to making a consistent profit, not Technical Analysis and not other
factors in trading systems, no matter how good such systems may or may not
otherwise be.
Technical analysis (a trading system) can provide 50-50 or somewhat better
indications of a price move, within some considered time period, plus some
indications of where to put stop losses and profitable target exits.
Statistically one can determine, sort of, that a particular set of
conditions based on chart patterns, candlestick patterns,
indicator/oscillator conditions, perhaps even some specific fundamentals,
tend to predict correctly some percentage of the time, within a range of
particular market conditions. (The past may have little to nothing to do
with the current or future, though, something that is weighed too lightly by
many in their back testing results.)
If one trades only with reward:risk ratios of 2:1 or better, limits risk to
between 0.4% to 1% of capital in any one trade, and sticks to the preplanned
stop losses, then any TA method that is 40% or greater in forecasting
accuracy, should, over time and over many trades, return moderate to large
profits.
If one does that, then it does not matter whether or not the markets are
random. Yes, draw downs might be large and consecutive losing trades might
be more than ten occasionally.
So, all the efforts to find a trading system with 90% winning trades, no
more than 3 consecutive losing trades, little to no draw down - all well and
good?
Well, maybe not so good. Traders would be far more successful accepting a
mediocre trading system and spending greater efforts at money and risk
management and self-discipline.
- - -
My short opinion is that the markets can appear to be random, because of the
incredibly complicated mix of nonrandom and random effects.
But who cares about the academic argument over how much randomness is in the
markets?
The Holy Grail of trading consists of knowing and trusting oneself well
leading to personal trading discipline; money and risk management; and a
mediocre or better set of methods, tools and infrastructure.
Vince Heiker
Flower Mound, Texas
---
|