Examining Seasonal Tendencies

November 2001 CSI Technical Journal
Holiday Schedule
CSI will be closed for voice communication on Thursday, November 22nd for the Thanksgiving holiday.
US exchanges will be closed, but data from other exchanges will be available as usual. The CSI staff
wishes you a Happy Thanksgiving!
Examining Seasonal Tendencies
To Uncover Clues in Historical Data Reserves
In these days of extraordinary uncertainty over earnings, energy prices and the economy in
general, many investors are searching for viable alternatives to the everybody-wins-or-everybody-loses
stock market. Although I am not convinced that fleeing the stock market now is necessarily a good
strategy, any time is a good time to consider investing in the traditional commodity markets, which are
now seen to be no more risky than stocks. These markets, which include grains, meats, metals, textiles
and energy products, always produce balanced gains and losses. Given these 50-50 odds, an average
trader might be on the right side of the market roughly half the time.*
Good timing is the key to success in commodity or stock transactions, and Unfair Advantage™
(UA) offers several tools to aid in determining the right time and direction for your trades. Among the
most important is CSI's exclusive Seasonal Index study. As the name implies, the Seasonal Index study
measures and reports seasonal tendencies for each market. These tendencies have an immense impact on
market prices, yet they are often underestimated or ignored. The Seasonal Index study is a quick and easy
way to gauge seasonal behavior for any given stock or futures market.
With a finely tuned index reading for each of the 251 trading days per year, CSI's Seasonal Index
illuminates the seasonal tendencies for each individual commodity or stock as no other tool can.
Traditional attempts at preparing seasonal indices are contrived by tracking average price behavior in a
monthly frequency. Most seasonal studies use the average price for each month over all years of history,
creating a very coarse monthly average that stands alone as a measure of seasonal tendencies.
Conversely, CSI's carefully designed cumulative index incorporates all of the historical data for a
given market into a daily time series that can be used to precisely gauge seasonal tendency on a daily
basis. UA’s seasonal study is perhaps 21 times more refined than the rather crude monthly data approach
adopted by other seasonal analysis techniques.
The operative term above is “cumulative.” The cumulative effect is that for each new day of data
introduced after the first full year, the seasonal study gains accuracy and definition. The inclusion of
successive days adds more confidence for each precise relative data point in time. As can be seen from
careful examination of a series like CSI’s #89, Heating Oil (see Figure 1), or a stock (see Figure 2) the
cumulative Seasonal Index paints an ever more refined index as more and more data are introduced.
Oftentimes, that index becomes predictive of the given market as a whole. The seasonal index's wave
form for the forthcoming one-year period is the same wave form painted over the most recent 251-day
trading period.
We could have overlayed the final year's seasonal waveform repetitively over all years of any
series instead of presenting the day-by-day cumulative result, but this, of course, would have been
misleading. The final year's seasonal waveform is uniformly derived from all of the data that precedes it."
The nearest-future chart of Heating Oil shows seasonal activity that tends to suggest long
positions near the end of the first quarter and short positions near the end of the third quarter. It appears
that Heating Oil does demonstrate some seasonal tendencies. These are generally attributed to demand
changes due to weather.
The computational algorithm for UA’s Seasonal Index employs statistics to calculate the chart
readings that are defined in relative standard deviation units, not absolute price changes. The index
represents a + or - 3 sigma confidence interval over time. Some Seasonal Index studies look like mirror
images of the recent past, while others reflect only subtle seasonal effects.
Figure 3 - A seasonal study of ToysRUs stock.
Seasonal Indices offer a way to combine seasonal information on market data with daily chart
analysis to promote a better understanding of future price movement. The seasonal wave can be used as
an input to forecast future events or to simulate past events without bias where any then-current reading
does not and cannot affect earlier seasonal patterns.
One way to use the Seasonal Index study is to look for coincident peaks and troughs between the
market and the index. When peaks coincide or when troughs coincide, the risk of loss on a trading
position should be at its minimum. For some traders this may translate into an opportunity to take a
heavier position. It is not recommended that seasonal indices be used as the sole basis for trading. They
should only be used in conjunction with other confirming market factors.
To add the Seasonal Index to a given commodity chart, first make a computed, continuous chart
of the underlying market. A back-adjusted or nearest-future series is recommended. Next click the
"Change Indicator on Chart" button on the toolbar. Click the "Set Study" button and then click the arrow
to display the drop-down study menu. Select "Seasonal Index" and then [OK].
Results may be improved by selecting Unfair Advantage's "detrend" option in creating the input
continuous series. This will remove any directional bias that may be present in the market to be studied.
Your computed contract must not contain any negative values.
We recommend this study be performed on at least ten years of daily data (much more if
possible). When years of substantial history are supplied, the cumulative general waveform becomes
more and more characteristic of the given market. You should not attempt a Seasonal Index calculation
with any time series holding less than one year of data.
Aside from controlling the input data set by selecting the continuous series of your choice, the
only user-defined parameter in the Seasonal Index computation is amplification. Amplification expands
the seasonal waveform, making it easier to visualize seasonal effects. Keep in mind that amplification
may distort the significance of seasonal effects because the actual measurement of the sigma readings will
be lost. Figure 3 shows the same seasonal study in amplified form (upper chart) and non-amplified form
(lower chart.)
This back-adjusted, detrended Corn data with CSI's Seasonal Index study reaches the full range of
from -3 to +3 sigma in the amplified study. The non-amplified standard Seasonal Index study uses the
more modest scale at left. Note that the wave form extends one year beyond the present date.
The Seasonal Index study supplies a logical methodology to graphically reveal seasonal
tendencies. Seasonal effects might reflect the ebb and flow of the growing season or cycles of demand
based on temperature shifts that go hand in hand with the changing seasons. In some cases, seasonal
effects reflect demand based upon cultural practices like holiday gift-giving and summer travel.
Trading decisions based on seasonality rely on the assumption that seasonal influences do, in fact,
affect market prices in predictable ways. Although the Seasonal Index can be appropriately used in testing
stocks for seasonal effects, it should not be used to look for clues to behavior of stocks that are clearly not
affected by seasonal fundamentals. When no seasonal correlation exists, you will not be able to find a
correspondence between the Seasonal Index and the market data. In these cases, the study will simply
reflect upon the apparent noisy characteristics of the data. Seasonal regularities in the underlying
fundamentals are only one factor in determining price, and seasonal trading should only be used to detect
seasonal tendencies if consistently applied over many years.
In these times when enthusiasm for the stock market has dampened, it may be a good time to
explore traditional commodity markets and highly seasonal stocks. These tend to retain their ability to
reward the savvy investor in good times and bad.
The examples shown were casually chosen from the thousands of markets in UA’s inventory.
Seasonal Indices derived from seasonal crops are probably the best candidates to exhibit reliable seasonal
patterns. This study thrives on abundant historical reserves, bringing to mind Patrick Henry’s famous
quotation, “I know of no way of judging the future, but by the past.” We wish you the best of luck with
Bob Pelletier
Important Notice: The information provided here is general and impersonal in nature, and is not
intended as trading advice. Readers are urged to exercise their own judgment and to consult professionals
familiar with your own situation before investing. Commodity markets are inherently risky, and only risk
capital should be used. Past performance does not guarantee future results.
Tech Talk
Each month in this column, the Technical Support Staff addresses topics of interest to many CSI
subscribers. This month they discuss software upgrades, and the effects of September 11 on the database.
Q. When I started using Unfair Advantage (UA), the installation process required that I store a series of
programs from ActiveState on my drive. As I recall, they had something to do with the Perl programming
language. I noticed that those screens were not included when my colleague began using UA a few days
ago. Why the difference? Can he still use Perl?
A. Not to worry! UA version 4.2 still supports Perl for custom programming through the charting module
and through MarketScanner.™ To ease installation, we now include a public domain version of Perl with
the UA executable installation. Your colleague's UA version sports a number of changes like this that
simplify various processes. Please visit the CSI website for more information on upgrading your
Q. How did the chaos of September 11th affect my UA database?
A. Despite the unprecedented events of September 11th, the American financial system performed
remarkably well and actual chaos was averted. Most exchanges were closed for just two days following
the terrorist attacks, but the U.S. stock exchanges did not reopen until September 17th. Each day of
closure is recorded as a "holiday" in your CSI data files. Those data sets that were missed due to late
postings in the aftermath have already been updated through UA's remote correction process. If you were
unable to download due to problems with your local server, all missing days were automatically supplied
upon your first successful access. We know of no latent problems that would require database
replacement or additional downloads.
Q. My UA stock charts show a gap in trading from September 11 through September 14th. While this
accurately reflects what happened, I wonder if I can display the data without a gap. Is there a way to hide
the break?
A. Not in your UA charts. However, when building an ASCII file, you'll have the choice of converting
"holiday" data into zeros, duplicates of the previous day or the previous close, or to simply reserve a null
day. These options might be useful for custom or third-party analysis.
Q. I'm looking at a price bar on a UA chart that is like none I have ever seen before. The "close" hash
mark is all by itself above the price bar. Is this an error?
A. Probably not, but please report any suspicious data point to our technical support staff
(techsupport@csidata.com) for confirmation. What you are seeing is more likely an out-of-range
settlement. An inherent trait of the commodity clearing process is that some inactive contracts (primarily
mercantile) are subject to settlement by committee -- sometimes at values outside of the day's trading
range. This allows the exchange to keep all contracts appropriately priced regardless of variations in
liquidity. Whenever a contract is "settled" outside of its trading range, an atypical price bar is produced
that includes a gap between the vertical range indicator and the closing tick. Many software programs,
UA included, can chart out-of-range closes (actually settlements) with no problem, but others cannot.
If your third-party analysis software can't handle out-of-range closes or if you do not wish to use this type
of data in your analysis, you can edit your portfolio settings to eliminate them. Here's how: First select the
desired portfolio from the Portfolio Manager panel. Then click the "Edit" portfolio button. Go to the
Rounding/Ranging tab and select from "As published" (can include out-of-range data), "Modify
High/Low" (expands the range to include the settlement) and "Modify Close" (adjusts the close to be
within the range). In all cases, the UA database retains the actual values. Your choice here affects only
data presentation for UA charts and export files listed in the selected portfolio. The same option is
available for Non-Portfolio Chart Options.
Q. The Columbus Day prices for many of my foreign currency contracts seem to be repeats of the
previous day's close. What's going on?
A. A procedural convention adopted by the Chicago Mercantile Exchange does indeed call for some
repeated prices on October 8th (Columbus Day). The affected contracts are the CME's foreign currency,
E-Mini and FX markets. The lead December 2001 contracts on Globex for the currencies and FX markets
traded, and in all or nearly all cases, prices reported were marginally higher. All further distant contracts
did not trade.
In the absence of market data for the farther out contracts, the exchange repeated the close from the
previous Friday as the official close for Monday. The volume reports shown for the distant contracts
beyond December for these markets were for "out-trades." The CME's official policy is to repeat the
settlement of the prior trading day, leaving us with no choice but to post the data as they released it.
UA customers can manually override the nominal settlements, but such an override could be lost when
installing a future new release.