|
Feature
 Understanding
Market Type Part III
by
Van K.
Tharp, Ph.D.
You
should always know how your trading system performs under various
market conditions. This
is the third article in our series on Understanding Market Type and
I’ve written several articles on the topic previously as well.
Last
week, my analysis suggested that none of the three market types we
had been looking at had much value in terms of weekly prediction;
thus, they are useless for determining whether a system will work in
that market type. One
reason why market type has such little predictive value during a 13
week period is illustrated by the results of our last analysis of
market type, published on May 24th, when all three market
types gave very different readings.
Method
2 classified the market as volatile bull, Method 3 classified the
market as volatile sideways, and Method 4 classified the market as
quiet bear. The market
had been going up enough to account for being positive over the last
13 weeks (but not when you avoid dropping the last down week 13
weeks ago as we do in Method 3); however, over the last 200 days
(Method 4), the market was still way down.
This resulted in mixed readings.
In
other words, you can have a bear market with huge up moves in the
last part of it, but not enough to make it net up.
Similarly, you can have a bull market with huge down moves in
the last part of it, but not enough to make it net down.
This means that market type has little predictive value.
In
our last ETF workshop, Ken Long mentioned how my discussion about
System Quality Number™ from a prior workshop got him thinking
about using SQN to screen the markets.
(What I love about our workshops is that the more you attend
them, the more good ideas you get.)
And from Ken’s discussion, I got another idea about how to
measure market type. What
if I looked at the SQNs of the daily percent change in the S&P
500 over 200, 100, 50, and 25 days?
That might be an excellent measure of market type.
So
we programmed XLQ to gather as much S&P 500 daily data as is
available and looked at the SQNs of the daily percent changes to
determine market type. I
used four different time periods to determine the differences:
200 days, 100 days, 50 days, and 25 days.
This
first graph
shows a histogram of the distribution of SQNs™
for 200 days. We had
14,743 days in our calculations.
The
average SQN was 0.687 and the median was 0.74.
71% of the 200 day SQNs were positive.
Let’s
now look at the 100 day SQN in the next graph.
Here the average SQN is 0.534 and the median is 0.51.
67% of the 100 day SQNs were positive.
The
next graph shows the distribution for 50 day SQNs.
Now
the mean is 0.422, the median is 0.37, and 64% of the SQNs are
positive. And for the 25
day SQNs, shown below, the mean is 0.334, the median is 0.31 and 60%
of the SQNs are positive.
Thus,
the shorter the time frame, the more likely the SQN is to be
negative. This clearly
shows the overall long term bias of the market in an upward
direction.
Based
upon these data, I decided that I would call an SQN of 1.5 or
greater extremely bullish. That
would be no more than 30% of the distribution.
I then decided that anything between 0.3 and 1.5 would be
bullish. Anything
between 0.3 and minus 0.3 would be a neutral market.
Anything between minus 0.3 and minus 1.0 would be bearish.
And anything less than minus 1.0 would be very bearish.
I then looked at frequency distributions for four SQNs, for
the average percent change for days falling within each category,
and for the average percent change for the following day.
These data are shown below.
Notice
that for every SQN grouping, the average percent change gets smaller
(or more negative) as we go from strong bull to strong bear.
Clearly we’ve done a fairly good job of distinguishing the
categories. But look at
the average change the day after.
Except for the 50 day SQN, the average percent change the day
after gets progressively smaller (or negative) as we move from
strongly bullish to bearish. In
addition, another interesting phenomenon occurs:
when the market becomes strongly bearish, we have the largest
percent changes the day after. In
fact, except for the 25 day, the largest day after average percent
change seems to occur when the market type is strongly
bear—amounting to over 0.5%.
Thus,
the SQN market types clearly describe the market as we’d like and
they all seem to have some predictive value.
As a result, we will start using the 25 and 100 day SQNs to
describe short and longer term market types in the future.
Next
week, we’ll continue our discussion of market type with a
discussion of volatility.
All
of this work was done with the XLQ add-on to Excel with the help of
Leo van Rijswijk, the developer of XLQ.
About
Van Tharp: Trading coach, and author, Dr. Van K. Tharp is
widely recognized for his best-selling books and his outstanding
Peak Performance Home Study program - a highly regarded classic
that is suitable for all levels of traders and investors. You can
learn more about Van Tharp at www.iitm.com.
|
