OBJECTIVE
BACKGROUND
METHODOLOGY
BRIEF INSTRUCTION
CUSTOMER INFORMATION
PERFORMANCE DATA FOR SELECTED STOCKS
TECHNICAL ANALYSIS WEBSITES
READ HOW QUANTITATIVE INVESTORS USE
THEIR MATHEMATICAL GENIUS TO OUTSMART THE MARKET
To
inform our clients and prospective investors on the basic
methodology used in our advanced technical analysis and thereby
reveal the overall methodology and performance of the MASTERStockAnalysis program.
The MASTERStockAnalysis is based on
a proprietary logic discovered by the author (Dr. Chehroudi)
after many years of intensive research work on technical
analysis of stocks. The program optimizes 8 to 10 key
parameters to provide the best performance (gain and/or
reward-to-risk ratio) during its operation. Re-optimization
is done in a period of about 30 days to be used for the next
period. Independent optimization and re-optimization is
performed for the short logic at the end of each period.
Any stock analysis and
recommendation program must deal with two issues. First, to
provide the buy/sell timing points, and second, to suggest a
buy/sell price when a trade timing is triggered.
MASTERStockAnalysis program has these two critical
components and uses both statistical and pattern recognition
algorithms to achieve them. However, there is one
newly-discovered and proprietary core buy/sell decision
making logic which is responsible for more than 80% of the
buy/sell signal points. To the best of the author's
knowledge, there is no place in the open literature where
this proprietary logic is discussed or even mentioned.
Therefore, the author confidently claims this to be an
original contribution. The novelty of it to the readers or
users, of course, can be truly proven only by monitoring its
performance for a sufficiently long period of time (2 to 3
months) or a number of trades (5 to 7). There are many
(over 80 or so) technical analysis approaches and indicators
in open literature to guide the buy/sell timing points. Out
of these only a few has been shown to provide statistically
significant performance than the passive "buy-and-hold"
strategy over the years, see Colby and Meyers[1]. By far
the most comprehensive published work this author is aware
of is the systematic work done by Colby and Meyers. Rigorous
testing of the MASTERStockAnalysis program includes many of
the methodologies discussed in this book. However, as will
be seen, the time scale for the application of the
MASTERStockAnalysis is tailored to be shorter than those
discussed by Colby and Meyers. It should be mentioned that
the MASTERStockAnalysis program is not designed for
day-traders although such a potential exists and can be
incorporated.
Almost all well-performed
programs must optimize their parameters using the past
historical data. This by itself should be optimized. That
is, one must find the best historical data period for the
optimization process to provide the highest performance.
Although some variations exists, it appears that a 6-month
period is the best for the MASTERStockAnalysis program in
most cases. The optimization efforts themselves are
controversial. Critics point out that future market
behaviors and price patterns may not follow those on the
past. Proponents respond that there is sufficient similarity
to make the optimization worthwhile. The correct method to
test technical decision rules is blind simulation or
ex-ante cross validation. Optimized rules should be
accurately defined on past historical data, then tested on
more recent data not included in the prior
optimization search. Only this way, one can gain a
realistic perspective on how it would be to project
past-optimized rules forward through time, just like in real
time. According to Colby and Meyers, this testing process
imposes the rigor of the scientific method on technical
analysis theories. Many studies suggest that the optimized
timing decision rules can show reasonably worthwhile results
when projected forward through real-time simulation.
However, one should keep in mind that because the future is
always somewhat different from the past, it is unwise to
expect future results to be as perfect as past optimized
performance. Nevertheless, experts believe that
optimization profitability testing offers a reasonable
approach to selecting effective decision rules and
practical investment strategy.
As indicated,
optimization process for the MASTERStockAnalysis program
uses 6 months of daily historical data for analysis
purposes. One can think of the optimization process as a
psychologist or a therapist accompanying someone on a daily
basis for a period of 6 months during which the analyst
observes and analyzes his/her behavior. Then after the end
of this period a good therapist would be in a well-informed
position to predict the person's response to life events
with higher averaged likelihood of success than anyone
else. For the MASTERStockAnalysis program, once determined,
the optimized parameters are held constant for the next two
to four weeks after each optimization event depending on the
market and stock performance. In principle, given a high
speed computer, the optimization can be done on a daily
basis. However, such a short time activity would be quite
unnecessary. Results on application of the
MASTERStockAnalysis for a selected stock (AOL) will be
shown next as an example. The AOL stock was purposely
chosen because of its poor performance in the past year. In
this case, a six-month period of the past historical daily
data from Nov, 9, 1998 was the starting point. The
MASTERStockAnalysis was optimized for this period and then
applied for the buy/sell decisions during the following
month. Of course in this case the next month's actual daily
price data is available and can be tested. This process
continued with monthly re-optimization for one year
and the results were compared with several strategies
including the "buy-and-hold" approach. Figure 1 shows the
buy/sell pattern exercised by the MASTERStockAnalysis
program for the AOL's stock for a period of one year.
The
red curve shows daily historical close price data and
the blue indicates the buy/sell
points. Horizontal segments of a line indicate periods
when program stayed out of the market (cash position). The
scale for the blue curve is on the right edge with the unit
of $/share and is 1/4 of that shown at the left edge of the
plot which is for the daily close price values. The
horizontal axis shows the time scale. At the end of the
one year examination period the gain achieved by the
MASTERStockAnalysis program was 20.68 $/share
whereas a posteriori optimization of this one year
actual simulated period gave 47.25 $/share.
A posteriori optimization refers to a situation where the
future historical data is somehow (God-ly) known for one
year and one uses the MASTERStockAnalysis program to
achieve the best performance possible. This is the maximum
gain the program would have delivered, if it had a crystal
ball and knew the future. It is important to note, and it
should be obvious, that in actual application the monthly
optimization simulation test does not see and know anything
about the future and therefore must consistently (on monthly
bases in this test) re-optimize as time progresses. This is
necessary in order to include the new data in each
optimization event. For this one-year test period, the
passive buy-and-hold produced -20.43 $/share
loss assuming that the buy and the sell events corresponded
to the program's first buy and the last sell points,
respectively. For the "buy-same/sell-max strategy," in
which the buy is assumed to be the same as the program's
first buy but the sell point be whenever the absolute
maximum of the daily price occurs, a 14.7 $/share
gain is possible. The "buy-min/sell-max strategy," being
realistic only if the maximum price occurs after the
minimum in time, is 50.44 $/share.
However, this strategy cannot be applied to AOL as the
minimum price occurs after the maximum in time. For those
stocks that this strategy is a possibility, results are
always considered and compared with that of the
MASTERStockAnalysis program. Finally, if one visually
inspects the same one-year daily historical data and selects
the best buy and sell points, then an absolute "ideal or
theoretical gain" of about 74.8 $/share
can be calculated. Obviously in real application such a
future information is not available and this exercise is
merely for determination of the theoretical maximum
gain for this one-year period irrespective of the
choice of the technical analysis program. To be
correct, this theoretical maximum gain of 74.8 $/share
should be compared with the 47.25 $/share when the
MASTERStockAnalysis program somehow (God-ly) knows the
future.
It is important to
indicate that in many cases it is not desirable to optimize
the gain, rather it is best to work with the
so-called reward/risk ratio or both. Reward/risk
ratio is defined as the "total gain for each 6-month
optimization period" divided by "the maximum single-trade
loss" for the same period. Other definitions of this ratio
such as "maximum single-trade gain" divided by "the maximum
single-trade loss" or "total gain" divided by the "total
loss" for each optimization period are also used to reach
to the final optimization parameters at each stage. During
each optimization exercise both of these performance
parameters (i.e. gain and reward/risk ratio) along with
the percentage of the positive-gain trades are closely
monitored for determination of the optimum program
parameters to be considered for the next two to four week
trading activities. It is worthwhile to point out that the
stock market is a highly nonlinear system with degrees of
randomness and fractal behavior and often has multiple
solutions. Many optimization routines have difficulties
finding the true optimized parameters (based on the gain,
reward/risk and % positive-gain trades performance criteria
discussed above) for the MASTERStockAnalysis program. The
author has developed a methodology to unequivocally identify
the optimized points.
The short/cover-short
logic of the program also goes through a similar rigorous
back testing and optimization procedure prior to its actual
application.
It must be emphasized
here again that the MASTERStockAnalysis program is not
currently designed for day-trading activities. A typical
scenario for its operation is as follows. Every evening
after the market is closed, the user initiates an access to
the YAHOO site to download daily historical data of the
desired stocks for a period of six months which includes
today's trading information. The user must ensure that
today's price information is indeed included in each file.
Each file is then imported into the program. Then the
MASTERStockAnalysis program issues either a buy (a jump
down for a long and a jump up for the short strategies, see
the Brief Instruction section below), a sell (a jump up for
the long and a jump down for the short), or a hold ( no
jump) signal. If a buy or sell is initiated, then the
program suggests the best price to be used in the online
broker software for the next day. This price is also
optimized so that, first, it is highly likely to be
materialized (i.e. be in between the LO and HI prices of the
next day) and, second, be the best value for the type of the
trade recommended (i.e. be closer to the LO if buying and to
HI if selling). For those situations where the suggested
price can not be materialized, the only remedy is through
user's midday inspection of the price activity the next day.
Note that the idea and the reason behind this approach is to
avoid user's continuous and full-time inspection of the
market as is often done by day-traders. One can have
other dedicated professional activities during the day and
combine it with a minimal stock market effort time in the
evening using the MASTERStockAnalysis for optimum
performance. However, there are cases where higher gain
is achieved if one can determine the buy/sell timing before
the market is closed rather than discovering it in the
evening after the historical data is downloaded. This
desired feature is available in the MASTERStockAnalysis
program. Every day the user should visit the online market
for collection of the "midday" stock activity information.
It is then manually imported to the program for sensitivity
analysis. If indeed a buy/sell signal is recommended, the
user should act before the market is closed. Although better
performance can be achieved in this manner, it requires one
midday visit of the market price activities for the stocks
of interest.
It is clear that the
proprietary decision rule in the MASTERStockAnalysis
program when optimized monthly using the past 6-month
historical daily data provided a robust and superb
performance considering that the AOL stock in this one-year
period performed very poorly.
It is hoped that by the
aforementioned explanation of the methodology and sample
results we have addressed some of the questions our clients
have posed to us.
[1] Colby, R. W, Meyers, T. A, The Encyclopedia of
Technical Market Indicators, McGraw-Hill, 1988.
Back to top
BRIEF INSTRUCTION FOR THE
"MASTERStockAnalysis" program
The thick black
square-curve indicator shows the buy/sell timing. A sharp
drop or jump down on this curve indicates a buy signal. The
size of the jump is equal to the suggested buy price. A
sharp rise or a jump up is a sell signal. Likewise, its size
is the suggested sell price. The cumulative gain after each
buy/sell cycle is the amount that this black indicator is
above the zero level. The scale for the buy/sell indicator
curve is shown at the left edge of the plot. Note that the
scale for the stock price may be logarithmic and
shown at the right edge of the plot. Therefore, care should
be exercised by the users in interpreting and comparing the
gain/loss shown for each trade cycle with that measured on
the price curve. The period between each vertical scale
lines (dashed or solid) is 4 days. It is important to
emphasize that even though moving averages are shown in the
plots the core of the decision makings (nearly 80%), which
is proprietary, is NOT based on the simplistic and classical
approach of using moving average cross-over rules.
The thick blue indicator shows the short (jump up) and
cover-short (jump down) points. The cumulative gain after
each short/cover-short cycle is the lower part of this thick
blue indicator.
CUSTOMER INFORMATION
Visitors interested in this
technology can sign a contract to receive emails at the
buy/sell and/or short/cover-short timings along with the optimized
suggested buy/sell prices for a nominal fee. Also, a client
version of the program can be purchased to be used in
conjunction with our monthly-updated optimized parameters as
inputs for a fee. Please
contact Advanced Technology
Consultants for more
details.
Back to top
-
Note that the scale for the stock price may be
logarithmic and shown at the right edge of the
plot. Therefore, care should be exercised in
interpreting and comparing the gain/loss shown for
each trade cycle with that measured on the price
curve.
-
The total $ gained should be calculated as the
sum of the gains for the long and short strategies.
Note that there are occasions that the short
strategy delivers a better performance than the long
strategy and vice versa. Exercising both strategies
increases the total $ gain in almost all cases.
-
Note, it is critical to recognize that after three
to four weeks from the last optimization date
(given for each stock below) the performance is not
expected to stay as high as before and a
re-optimization is to be performed.
-
As indicated in the above, higher gain than
shown below is achievable if one can determine the
buy/sell short/cover-short timings before market is
closed rather than discovering it in the evening
after the historical data is downloaded. This
feature is available but not used for the trade
decisions of the selected stocks given below.
-
Following results are not updated daily and shown
here solely for comparison and program performance
purposes.
AOL:
LAST POSTED OPTIMIZATION UPDATE:
NOK:
LAST POSTED OPTIMIZATION UPDATE:
QQQ
LAST POSTED OPTIMIZATION UPDATE:
Back to top
TECHNICAL ANALYSIS
WEBSITES:
READ
HOW QUANTITATIVE INVESTORS USE THEIR MATHEMATICAL SKILLS
Wall Street Rocket Scientist
Stock Picking Computers
Back to Top of This Page
NOTE:
Contact Advanced Technology
Consultants for consulting needs and opportunities in this
area.
