1
Active Management of the Rydex Sector Funds
Abstract
The paper presents a simple method of active management of the Rydexsector mutual funds. It is a group of sixteen funds that began trading in 1998, andlike most Rydex funds, there are no restrictions on the frequency of trading or anypenalty charges for short holding periods. The underlying concept is to buy thefund that recently has been the weakest performer of the group using a standardtechnical indicator, the Commodity Channel Index (CCI), hold it for a few days,and then exchange to the fund that has become the weakest. The steps of thedevelopment of the trading method are shown, which includes a mechanism tocontrol risks based on the system’s equity curve to indicate when to stop and thenresume trading. Since the current secular bear market started in 2000 and theRydex sector funds have existed since 1998, the analysis and system developmentare based on 2000 through 2010.
The first step demonstrates that buying the sector fund with the lowest CCIis a viable concept by comparison to owning the one with the highest CCI. Nextthe first two parameters are found. They are a 16 day moving average in the CCIcalculation and a three day minimum holding period for the fund purchased. Theseare determined simultaneously using the Ulcer Performance Index (UPI) as themeasure of merit. The performance of the basic system using these parameters isexamined, and while the investment return is outstanding, the drawdowns are toolarge for most traders to accept.
As a method to control risk, we find when to cease trading the system usinga simple moving average crossover based on the basic system’s equity curve. Theparameters of the crossover are 20 days for the shorter one and 100 days for thelonger one. These values are determined holding the basic system parameters (16and 3 as explained above) constant. In effect, the moving average crossover servesas a filter for the basic trading system.
The performances of the active management methods are compared to fourmethods of passive management—an index fund, the daily average performance ofall the sector funds, passive asset allocation with the benefit of hindsight, owningthe best performing sector fund each year—and shown to be superior to all but thelast one, which requires perfect forecasts of the best performing Rydex sector fundfor eleven consecutive years.
A compelling argument for the wisdom of active management results fromthe far superior investment returns and much lower risk levels of the tradingsystem in comparison to passive asset allocation, even when it has the benefit of
2
hindsight. For that test, four not highly correlated sector funds with the highestreturns over 2000-10 were assumed to have equal dollar amounts invested in themat the start of 2000 and rebalanced to equal dollar amounts at the end of each year.Such a process is typical of the investment advice given by “mainstream experts.”
The trading system is virtually equal in performance to owning the bestperforming sector fund each year, which is another demonstration of the value ofactive management. While no person or system is going to pick the sector fundwith the highest return every year or even most years, using a fairly simple tradingsystem can result in comparable levels of return and risk. In this sense, activemanagement can make up for the lack of a crystal ball.
Next, some additional topics are discussed. One concerns how muchconfidence we can have that the trading method, which was created by backtesting2000-2010 data, will perform well in real time. While it is too soon to make adetermination, for six years the author has traded the Rydex sector funds using amore complex system based on buying the one with the lowest CCI and holding itfor a minimum of a few days. The results have been quite satisfactory.
Two other ways of system development are discussed next: walk-forwardtesting and trying to “optimize” the four of the parameters mentioned above alltogether rather than two at a time as was done. Neither of these was pursued, andthe primary reasons are discussed.
Another topic is an exercise to demonstrate the utility of active managementand the applicability of the trading method to a different asset class. Thedevelopment process employed to determine the trading system for the sectorfunds is applied to a group of single country exchange traded funds. The resultingtrading system with the same structure, but different parameters, outperforms thetwo passive methods used for comparison by a wide margin.
1
Introduction
After the debacle in stock prices in 2007-09 that saw the S&P 500 Index fall
by over 50%, all NAAIM members and a large portion of the investment
management community know that active management of portfolios is far superior
to “buy and hope” even with tweaks such as periodic rebalancing. There are quite a
few effective methods of active management. Perhaps the most popular one and
the one most written about is trend following. Often that method is combined with
another idea to make it even more effective. One notable example appears in the
winning paper of the first Wagner competition. It enhances trend following with
the addition of mean reversion.
It is worth noting that there are many ways to “skin the cat” using active
management. This paper presents one that is quite different from trend following
and applies it to the Rydex sector funds. It shows that there is no “magic formula”
when it comes to active management; there are many approaches that will work if
they have sensible formulations and are carefully researched and tested.
The essential idea behind the approach presented is that a sector fund that
has recently been the weakest performer in a group of them may well be
temporarily out of favor or oversold in comparison to the others in the group. If so,
owning such a fund for a few days can take advantage of a likely rebound and the
expectation that it will do better over the period than the typical sector fund in the
group.
2
Sector funds are not diversified, so they usually are more volatile than broad
market measures such as the S&P 500. If the fund is gaining, the higher volatility
should result in greater profits than the broad market. However, during weak
markets, sector funds tend to fall more than the popular indices. Moreover, the
weakest sector funds are quite likely to suffer substantial and painful gains when
stocks are trending down. Accordingly, the general approach described above calls
for a risk control mechanism to prevent drawdowns that are too large to make it
practical no matter what the level of longer term investment returns.
There are many, many technical indicators that can be applied to decide
when and in what direction to trade. The book, The Encyclopedia of Technical
Market Indicators by Robert Colby (McGraw Hill, 2003) contains almost 800
pages in its comprehensive listing. The author of this paper did not try to test a
variety of them. Rather one that “felt right” for the approach described above was
chosen and tested. As it turned out, using the Commodity Channel Index (CCI) led
to a trading system the author found to be quite desirable. Another standard
indicator, a moving average crossover applied to the system’s equity curve, proved
to be an effective risk control method.
3
Preliminary Topics: Rydex Sector Funds, Commodity Channel Index
There would be little point in trying to apply the approach described above
on a group of funds with similar holdings. What is desired is a set in which some
of the funds have the potential to perform quite differently at times than most of
the others. With the anticipated short holding periods, being able to move from one
fund to another with minimal or no cost is another needed feature. The latter
essentially requires using funds in the same mutual fund family that can be traded
frequently without costs or penalties. One of the few groups that fits the
description is Rydex sector funds. Table 1 lists the ones used here.
There are two other Rydex
funds that could be considered
sector funds: Precious Metals
(RYPMX) and Real Estate
(RYHRX). They were not included
because they are not sectors of the
economy in the sense the others
are. In addition to the sixteen funds
in the table, the Rydex U.S.
Government Money Market Fund (RYMXX) was in the database for the times
when no sector fund is owned.
Despite its name, the Commodity Channel Index (CCI) is applicable to any
tradable security. Its typical application as described in the Colby book (p. 155-
Fund Start Date Ticker
Banking April 1, 1998 RYKIXBasic Materials April 1, 1998 RYBIX
Biotechnology April 1, 1998 RYOIX
Consumer Products July 6, 1998 RYCIXElectronics April 1, 1998 RYSIX
Energy April 21, 1998 RYEIXEnergy Services April 1, 1998 RYVIX
Financial Services April 1, 1998 RYFIXHealth Care April 17, 1998 RYHIX
Internet April 5, 2000 RYIIXLeisure April 1, 1998 RYLIX
Retailing April 1, 1998 RYRIXTechnology April 14, 1998 RYTIX
Telecommunications April 1, 1998 RYMIXTransport April 2, 1998 RYPIXUtilities April 4, 2000 RYUIX
TABLE 1
RYDEX SECTOR FUNDS
4
158) is as a momentum indicator designed into trade breakouts to periods of
expected strong or weak performance. The approach here uses it in a quite
different manner: to identify the sector fund that has been the weakest recently as
indicated by the lowest CCI reading among the group.
The formula: CCI(n) = (P – MA(n))/(0.015*D)
where P = price of the fund1
MA(n) = simple moving average of last n days’ pricesD = mean of the absolute values of (P-MA(n))0.015 is a standard scaling factor that does not affect the methodpresented here
According to Colby, the normal use is to buy long when CCI exceeds 100 and sell
when it falls below that level; sell short when it falls below -100 and cover when it
exceeds that level. The 0.015 is needed to make those values effective. It has no
effect on which fund has the lowest CCI at any time.
The concept behind the CCI is determining when the instrument is making
an unusually large move, in either direction, away from its average behavior
relative to the moving average. A reading above 100 or below -100 indicates
positive or negative momentum that is increasing in the indicated direction. The
fund with the lowest CCI can be either the one with the greatest negative
momentum, when its CCI reading is less than zero, or the one with the weakest
positive momentum when all of the funds in the group have a positive CCI. In
either case, it has the weakest recent performance of the group as measured by the
indicator.
1 For a security traded during the day, instead of the closing price, the average of the high, low, and close wouldreplace P.
5
Trading System Development
Since regular mutual funds are priced once a day at the close, the
computation of the CCI and subsequent determination of the fund with the lowest
one can only be done after the market has closed. Consequently, any practical
trading system using the approach of this paper requires “next day trading.” In the
testing and reporting that follows, it is assumed that orders to buy, sell, or
exchange Rydex sector funds are executed at the NAVs on the market day
following the one that generated the signal.
As Table 1 shows, the first Rydex sector funds began trading in April 1998,
and by the end of that year there were 14. Two more were added in April 2000. We
are in a secular bear market that began in 2000, and the typical ones last 16-20
years, so the current one figures to be around for at least another five years. In a
secular bull market, buying and holding as well as the most aggressive forms of
trading and investing do well. It is in the secular bear markets that the benefits of
active management come to the fore. By controlling risk levels, it enables investors
to stick to their plans and maximize their chances of reaching their financial
objectives. With the secular bear market in mind, I decided to focus the system
development on 2000-2010. The returns and other measures shown below cover
that period.
The first thing to check is that buying the sector fund with the lowest CCI
has the potential to be the basis of a profitable trading system. Why not own the
one with the highest CCI? By that measure, it has the most upside momentum of
6
the group, so it might be a better one to own. Tables 2 and provide the answer.
They show the results of holding the Rydex sector fund with the lowest CCI based
on the previous days close, trading every day if necessary, and for holding the one
with the highest.
The left columns in the groups of three in each table show the length of the moving
average used in the CCI calculations. Higher values were also tested, but they did
not have returns or drawdowns as favorable as the better ones in the table. For each
moving average length, owning the lowest ranked fund by the CCI had a much
better return than the one with the greatest CCI. Except for a few with longer
lengths, the drawdowns were substantially less for the lowest CCI.
CCI Compound Maximum CCI Compound Maximum CCICompound Maximumlength (n) Return Drawdown length (n) Return Drawdown length (n) Return Drawdown
3 5.7% -65.7% 10 10.6% -72.0% 17 15.9% -60.4%4 3.1% -67.9% 11 11.5% -69.0% 18 16.9% -60.8%5 12.8% -44.2% 12 13.1% -63.5% 19 13.4% -71.5%
6 15.0% -53.4% 13 14.8% -56.9% 20 12.3% -72.3%7 11.4% -65.5% 14 10.8% -59.8% 21 8.9% -75.8%8 9.3% -67.5% 15 17.1% -52.5% 22 9.3% -74.8%
9 11.5% -67.9% 16 17.5% -61.0% 23 8.5% -71.3%
TABLE 2: 2000 - 2010 ANNUALIZED RETURNS AND MAXIMUM DRAWDOWNSWHEN HOLDING THE RYDEX SECTOR FUND WITH THE LOWEST CCI
CCI Compound Maximum CCI Compound Maximum CCICompound Maximumlength (n) Return Drawdown length (n) Return Drawdown length (n) Return Drawdown
3 -12.4% -91.9% 10 -4.2% -81.2% 17 -3.8% -81.3%4 -8.8% -85.9% 11 -1.0% -74.3% 18 -3.2% -76.2%5 -7.3% -86.5% 12 0.3% -80.0% 19 -1.8% -76.8%
6 -6.6% -83.9% 13 0.3% -77.1% 20 -2.6% -72.9%7 -3.4% -81.9% 14 -1.4% -80.7% 21 -0.1% -68.7%8 -4.6% -78.8% 15 -2.3% -80.4% 22 1.8% -63.1%
9 -6.0% -84.2% 16 -2.7% -78.5% 23 0.6% -67.8%
TABLE 3: 2000 - 2010 ANNUALIZED RETURNS AND MAXIMUM DRAWDOWNSWHEN HOLDING THE RYDEX SECTOR FUND WITH THE HIGHEST CCI
7
The drawdowns are uncomfortably high for all the cases shown. While the
tables favor owning a fund with the lowest CCI, perhaps selling it as soon as it is
no longer the lowest is not the best way to go. The thinking is that we are buying a
relatively week performing fund compared to the others in the group expecting it to
rebound, but that rebound may take a few days to be seen. It is worth testing to see
if imposing a minimum holding period can improve both the rate of return and
reduce the maximum drawdown.
An important question when developing a trading system is what “measure
of merit” to use when evaluating the possible methods. Naturally, we want higher
expected returns, but achieving those may entail levels of risk that are so large as
to make actually trading the system impractical. In other words, there is often a
trade-off between potential profits and risk levels. Using an evaluation measure
that takes into account both returns and risk levels is an excellent way to handle the
trade-offs.
There have been books written2 on how to develop trading systems and
choose an appropriate measure of merit. I decided to use a risk-adjusted return. The
return is the 2000-2010 annualized return, but what measure of risk is to be
incorporated? Maximum drawdown captures the worst case scenario, but is not the
best way to measure risk levels over an eleven year period. I decided to use the
Ulcer Index as the risk measure. It takes all drawdowns into account as well as
their prevalence. It is calculated as the square root of the average of the squared
2 Two are Howard Bandy’s Quantitative Trading Systems (Blue Owl Press, 2007) and Robert Pardo’s The Evolutionand Optimization of Trading Systems (Wiley, 2008).
8
periodic drawdowns, the period being one day in our case. Like the more common
standard deviation, which has a similar type of calculation, lower values indicate
less risk. Replacing standard deviation in the well-known Sharpe Ratio by the ulcer
index produces the Ulcer Performance Index (UPI), which is the measure of risk-
adjusted return that I will use in what follows.
Taking the above into account, the next step is to find the UPI for various
combinations of the CCI moving average length and the minimum holding period
(in days) once a sector fund with the lowest CCI is purchased. Table 4 shows the
results:
2 3 4 5 6 7 8 9 10
Length 3 0.36 1.53 0.29 0.29 0.14 -0.12 0.27 0.10 -0.09of 4 0.19 1.56 0.50 0.31 0.75 0.57 0.13 -0.12 0.03moving 5 0.94 1.43 0.24 -0.12 0.55 0.85 0.27 0.28 0.32
average 6 1.34 1.60 0.56 1.17 -0.09 0.45 0.24 0.20 0.35in CCI 7 0.24 0.52 0.62 1.18 0.13 -0.03 -0.03 0.39 0.32
(days) 8 0.11 0.92 0.89 -0.13 0.11 0.05 0.18 0.44 0.189 0.40 0.81 0.84 0.34 0.68 0.24 0.34 0.20 0.19
10 0.49 0.50 0.55 0.62 0.01 0.22 0.49 0.30 -0.06
11 0.90 0.89 0.59 0.18 0.36 0.23 0.18 0.24 -0.0412 0.83 1.19 0.47 -0.11 0.53 0.22 0.07 0.11 0.2513 1.01 0.55 0.24 -0.01 0.41 0.07 0.29 0.17 0.11
14 0.62 0.92 0.27 0.07 0.13 0.03 0.16 -0.12 0.0015 1.82 1.96 0.77 0.44 0.72 -0.11 0.17 -0.09 0.10
16 1.32 2.08 0.35 0.37 -0.02 0.00 0.16 -0.11 0.2117 1.22 2.10 0.48 0.21 0.07 0.05 0.06 0.01 0.2218 1.30 1.68 0.30 0.04 0.15 0.12 0.15 -0.02 0.17
19 0.67 0.35 0.17 0.14 0.26 0.09 0.12 -0.10 0.0020 0.52 0.07 0.24 0.22 0.13 0.02 -0.05 -0.08 -0.0121 0.35 0.17 0.17 0.15 0.00 0.05 -0.07 -0.06 -0.04
22 0.35 0.16 0.07 0.20 -0.03 -0.01 -0.11 -0.10 -0.1523 0.52 0.50 -0.03 0.01 -0.07 0.01 -0.14 -0.10 -0.15
24 0.55 0.37 -0.09 0.01 0.41 0.06 -0.07 -0.18 -0.0925 0.31 0.28 -0.02 -0.07 0.03 -0.04 -0.11 -0.20 0.06
Minimum holding period (in market days)
TABLE 4: ULCER PERFORMANCE INDEX WHEN OWNING FUND WITH LOWEST CCI
9
The boxed area is the “sweet spot” of the table. All of the values in it are
larger than most of the other entries. The drop-off once the minimum holding
period gets larger than three days is notable. The table is telling us that a moving
average of 15-18 days combined with a two or three day minimum is the best way
to go. I decided to use a 16-day moving average in the CCI and a three day
minimum holding period. Its UPI is slightly lower than the 17 and 3 value, but the
drop-off is larger going from 17 to 18 than for 16 to 15 days. I suspect that any
combination in the boxed area would lead to a trading model nearly as effective as
the one in this paper.
Buying the Rydex sector fund with the lowest CCI(16), holding it for at least
three market days and exchanging to the new lowest one after three or more days
(all trading on a next day basis) results in the following statistics for 2000-2010.
The compounded annual rate of return is 28.2%, the ulcer index is 12.6%, and the
UPI is 2.08. Even with those “nice” numbers, the maximum drawdown is 49.7%
on November 20, 2008. That value is too large for most traders, investors, and
money managers. One way to deal with it is to limit exposure to sector fund
trading, which is generally a good idea due to the volatility of such funds. Even if
the effect on the entire portfolio is not all that painful, it is extremely hard to stick
to a trading method that can suffer a hypothetical, backtested drawdown that large.
I will discuss how to reduce that measure of risk a little later. Before doing
so, let’s take a look at the annual performance of the system in order to get a better
feel for it. Table 5 shows the yearly returns and maximum drawdowns (starting
10
fresh at the beginning of each year). Since the Rydex sector funds began trading in
1998, I included 1999 in the table although it was not included in the system
development testing shown above. For comparison purposes, the table also shows
the comparable data for the Vanguard Index 500 fund (VFINX) that closely tracks
the S&P 500 with dividends reinvested and the daily average performance of all of
the Rydex sector funds. The latter is equivalent to a hypothetical fund that owns all
sixteen of the funds in equal dollar amounts and rebalances every day. It is also the
same as the statistically expected performance of owning a randomly chosen sector
fund each day. I include it to show that the performance of the system being
developed is not due to being “lucky” by choosing from an exceptionally strong
performance group. The top three rows show the compounded rate of return for
2000-10: 28.2% 0.3% 2.0%
Ulcer Ind: 12.6% 23.6% 22.8%
UPI: 2.08 -0.07 0.00
Return Max DD Return Max DD Return Max DD
1999 26.4% -23.9% 21.1% -11.9% 26.3% -11.8%
2000 92.9% -17.6% -9.1% -17.3% -4.4% -16.5%
2001 50.5% -35.7% -12.0% -33.5% -14.4% -29.2%
2002 48.7% -27.3% -22.1% -34.8% -21.8% -33.0%
2003 55.0% -11.5% 28.5% -12.9% 33.8% -13.8%
2004 25.5% -15.9% 10.7% -11.7% 12.6% -7.5%
2005 7.0% -14.6% 4.8% -7.9% 7.8% -7.0%
2006 5.8% -15.7% 15.6% -11.3% 11.4% -7.5%
2007 22.8% -10.9% 5.4% -10.5% 5.1% -9.9%
2008 -21.5% -49.7% -37.0% -49.0% -36.2% -47.7%
2009 27.8% -33.4% 26.5% -24.6% 35.6% -27.2%
2010 31.0% -13.8% 14.9% -16.3% 17.5% -15.7%
Avg. sector fund
TABLE 5: PERFORMANCE OF TRADING SYSTEM & "RIVALS"
Trading System VFINX
11
2000-2010, the ulcer index, and ulcer performance index3 for that period. Active
management using the trading system is far superior to “buy and hope” as
represented by the index fund. Not surprisingly, the average Rydex sector fund is
not much better than the index because as a group those funds encompass a broad
swatch of the market. In most years the trading system significantly outperformed
the other two. The year in which it was a laggard was 2006. Its only losing year
was 2008, which shows the market “crash” from mid-2007 until early 2009 were
of a different nature than the 2001-02 plunge.
Risk Control
Table 5 shows that the trading system is far superior to buy and hold as
represented by the Vanguard index fund or by owning an average or randomly
chosen Rydex sector fund. However, most money managers and individual traders
would not be willing to trade it as shown there. The problem is that there are
uncomfortably large drawdowns. The worst, in 2008, was nearly half the value of
the equity. Two other years, 2001 and 2009 saw drawdowns within the year of
more than a third of the assets.
I believe the most important purpose of active management is keeping risks
to levels that are acceptable. What is acceptable clearly varies by individual and
portfolio managers’ circumstances, but if the drawdowns are too large for comfort,
then the trading or investing method will be abandoned at least temporarily. Being
3 The risk-free rate of return for the period is that of the Rydex U.S. Government Money Market fund, which was2.0%. The T-Bill rate was a little higher, so using in the UPI calculations would have lowered them by a smallamount, but not affected the comparisons.
12
able to stay with one’s investment plans greatly increases the chances of achieving
financial goals. In short, active management can greatly increase the probabilities
of reaching financial goals.
There are a variety of risk control methods. One is using an external timing
model, of which there is no shortage, to decide when to follow the system and
when to move to cash equivalents. I do not think that is a good way with the lowest
CCI system because the system has the ability to do well even when the broad
market is faltering. As a simplistic example, suppose we had a crystal ball that
would tell us whether the S&P 500 was going to be up or down in the coming year.
If we applied its perfect predictions, as can be seen in Table 5 the CCI system
would not have been followed in 2000, 2001, 2002, and 2008. The last of these
was the only year the system lost money, and in the first three years it had huge
profits.
The method I prefer is to let the system itself tell me when I should stop
trading it and when I should resume trading. There are many ways that can be
done, and I examined only a few of them. My goal was to use one that was simple
and easy to understand. It is possible that a more “sophisticated” one compared to
what is shown below would be more effective. That research, either by the author
or someone else, can wait.
13
The method chosen is a simple moving average crossover based on the
trading system’s equity curve,4 which is computed even when actual trading has
stopped. In Table 6, the measure of merit is still the UPI and the parameters
determined above—16 day moving average in the CCI and three day minimum
holding period—do not change. In other words, the “optimization” of the moving
average crossover parameters is subsequent to and does not affect the
determination of the CCI trading ones. The boxed area contains most of the better
UPI readings although some of the ones in it are not particularly impressive. It
suggests that a relatively “slow” crossover method is best, with a short average of
4 The equity curve calculations start at the beginning of 1999 so there is a full year of initialization prior to themoving average testing.
5 10 15 20 25 30 35 40Length 5
of long 10 0.46moving 15 0.32 0.71average 20 1.23 1.02 0.91
25 1.45 1.21 0.77 0.7030 1.34 0.95 0.48 0.75 0.9835 1.65 0.77 0.95 1.31 0.94 0.90
40 1.44 0.84 1.27 1.28 1.24 0.79 1.2045 1.34 1.08 1.39 1.50 1.13 0.83 1.59 1.94
50 0.91 1.46 1.29 0.96 1.21 1.53 1.55 1.7360 1.22 1.30 1.51 1.18 1.27 1.29 1.31 1.8170 1.31 1.18 1.37 1.57 1.34 1.39 1.63 1.65
80 1.01 1.04 1.16 1.56 2.02 1.70 1.54 1.3790 1.22 1.12 1.26 2.21 2.12 1.60 1.12 1.15
100 1.33 1.29 1.66 2.58 1.95 1.26 1.28 1.47
110 1.11 1.35 2.04 2.23 1.19 1.06 1.27 1.35120 1.33 1.64 1.94 1.44 1.51 1.20 1.48 1.59
130 1.64 1.61 1.68 1.50 1.27 1.43 1.59 1.33140 1.44 1.38 1.68 1.75 1.41 1.49 1.40 1.56150 1.50 1.52 1.69 1.39 1.29 1.57 1.71 1.92
160 1.53 1.53 1.78 1.51 1.45 1.51 1.30 1.55170 1.33 1.51 1.56 1.60 1.52 1.40 1.12 1.42180 1.08 1.26 1.51 1.46 1.73 1.55 1.39 1.70
190 0.98 1.25 1.51 1.50 1.72 1.37 1.53 1.76200 1.12 1.50 1.55 1.55 1.49 1.31 1.59 2.26
TABLE 6: UPI WHEN TRADING SYSTEM ACCORDING TO MA CROSSOVERLength in days of short simple moving average
14
15-25 days and a long average of 80-120 days. I did not test intermediate values
not shown in the table (e.g. 17 and 93 days) because I would not have any more
confidence in such fully optimized parameters than the ones shown in the table.
Choosing the ones with the highest UPI, 20 days and 100 days, appears to be the
best choice. The UPI is much better, 2.58 vs. 2.08, than the one for the trading
method without the moving average crossover filter. Next, we look at the effects
on investment returns and drawdowns of the application of the MA crossover.
Table 7 below repeats data in Table 5 and adds information about the lowest
CCI system filtered by the 20/100 MA crossover (trade the system when the 20
day simple moving average of the equity curve is above the 100 day MA). It also
has two additional rows at the top showing the maximum drawdown in 2000-2010
and its date.
2000-10: 28.2% 0.3% 2.0% 26.7% 2000-10
Max DD: -49.7% -55.3% -53.3% -28.7% Average
date of: 11/20/08 3/9/09 3/9/09 9/26/01 Exposure
Ulcer Ind: 12.6% 23.6% 22.8% 9.3% 69.5%
UPI: 2.08 -0.07 0.00 2.58
Return MaxDD Return MaxDD Return Max DD Return Max DD Exposure
2000 92.9% -17.6% -9.1% -17.3% -4.4% -16.5% 92.9% -17.6% 100.0%
2001 50.5% -35.7% -12.0% -33.5% -14.4% -29.2% 32.2% -28.7% 83.1%
2002 48.7% -27.3% -22.1% -34.8% -21.8% -33.0% 43.3% -18.0% 70.6%
2003 55.0% -11.5% 28.5% -12.9% 33.8% -13.8% 60.6% -9.8% 91.3%
2004 25.5% -15.9% 10.7% -11.7% 12.6% -7.5% 20.8% -8.5% 44.4%
2005 7.0% -14.6% 4.8% -7.9% 7.8% -7.0% 8.3% -9.6% 68.3%
2006 5.8% -15.7% 15.6% -11.3% 11.4% -7.5% 10.2% -8.5% 57.8%
2007 22.8% -10.9% 5.4% -10.5% 5.1% -9.9% 11.1% -12.4% 74.5%
2008 -21.5% -49.7% -37.0% -49.0% -36.2% -47.7% -10.0% -13.5% 37.5%
2009 27.8% -33.4% 26.5% -24.6% 35.6% -27.2% 41.9% -13.8% 63.9%
2010 31.0% -13.8% 14.9% -16.3% 17.5% -15.7% 4.9% -21.8% 73.4%
Trading System VFINX Avg. sector fund System &MA filter
TABLE7: PERFORMANCEOF TRADINGSYSTEM, "RIVALS", & WITH 20/100 MAFILTER
15
Compared to the trading system without the filter, the system with the filter has a
slightly lower rate of return, 26.7% vs. 28.2%, but much lower risk levels. The
ulcer index is about one-quarter lower and the maximum drawdown is quite a bit
smaller, 28.7% vs. 49.7%. Moreover, the loss in the only losing year, 2008, is more
than cut in half. The downside of the filtering is evidenced by 2010 when the
filtered version of the trading system trailed the basic system by 26%. For 2000
through 2009 the two versions had virtually the same compounded rate of return,
near 28%.
As discussed above, I believe most would not be able to stick with the basic
system due to too many major drawdowns. My recommendation to anyone
wanting to implement the trading methods described here is either to use the
filtered version or to find another method of reducing the risk. To summarize, the
system to be implemented is:
Buy the Rydex sector fund with the lowest CCI(16) Hold it for a minimum of three market days Once another fund has the lowest CCI(16), exchange to that fund and
hold it for a minimum of three days Track the equity curve of the system
o When the 20 day simple MA crosses below the 100 day MA,move to cash
o Resume trading when the MA(20) crosses above MA(100)
16
The graph shows the unfiltered equity curve (black line) and the equity
curve with the 20/100 moving average crossover (green line). A portfolio
consisting of a single sector fund at any time is inherently volatile and is highly
unlikely to have a “smooth” equity curve. That is certainly so for the black line.
Using the moving average crossover as a filter, the green line, smoothes it to a
considerable extent. I consider it to be an appropriate way to trade sector funds in a
risk-tolerant account.
Figure 2 on the next page shows much of the data in Table 7 in a graphical
comparison of the annual performance of the trading system, filtered and
unfiltered, compared the Vanguard Index 500 fund. I believe it makes an extremely
strong case for the benefits of active management. In addition to the much lower
risk, the trading systems had only one losing year as compared to four for buy and
hold. In the years when the index fund had gains, they were generally less than
FIGURE 1: BASIC AND FILTERED (BY 20/100 MA CROSSOVER), 2000 - 2010(y-axis is log scale)
12/31/99 12/31/00 12/31/01 12/31/02 12/31/03 12/31/04 12/31/05 12/31/06 12/31/07 12/31/08 12/31/09 12/31/10
1000
4000
7000
10000
13000
Basic System
System filtered by MA crossovers
2500
17
those of the trading systems. Although not shown in the graph, the theoretical
equivalent to buying and holding a Rydex sector fund—owning equal dollar
amounts in all of them or owning a randomly chosen one each day—performs
similarly to the index fund.
Comparison to Passive Methods
The author has never put much stock (pun intended) into passive investment
methods. In particular, that means not collecting and saving articles about such
methods for owning sector funds. One approach that is commonly presented, for
more than just sector funds, is to select a few not highly correlated ones to own for
a long period of time with periodic rebalancing. Sometimes this method is called
passive asset allocation. I will examine how this might have worked for the Rydex
sector funds for 2000-10.
FIGURE 2: RETURNS, DRAWDOWNS FOR TRADING SYSTEMS, VANGUARD INDEX 500 FUND
-60%
-40%
-20%
0%
20%
40%
60%
80%
100%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Solid Bars - Annual ReturnsStriped Bars - Max DD in year
Green (left) - Filtered Trading SystemBlue (center) - Unfiltered Trading
SystemRed (right) - Vanguard Fund
18
Table 8 shows the annualized return and maximum drawdown for each of
the Rydex sector funds in 2000-2010. It shows the same data for the trading
system, without and with the moving
average filter. We see that none of the
individual funds had a rate of return
anything close to active management.
Only two of them, Consumer
Products and Health Care, had
maximum drawdowns that were
smaller than that of the unfiltered
trading system, and those two were
still greater than that of the filtered
trading system.
To formulate the passive asset allocation with rebalancing approach to report
on, I am going to “cheat” and pick four funds with the best returns over the period.
The best was Energy, and the second best was Energy Services, which won’t be
included because of its correlation to Energy. The third best was Basic Materials,
the fourth was Health Care, and the fifth was Consumer Products. Those three are
not highly correlated with each other, so I will include them in the four funds for
passive asset allocation. Note that the two with the smallest maximum drawdowns
are included. It is also worth noting that none of the high technology funds is
included. I suspect that most anyone picking four funds just before the start of
Fund Return Max DDBanking -1.6% -79.4%Basic Materials 6.6% -65.7%Biotechnology 1.5% -74.4%Consumer Products 4.2% -39.0%Electronics -7.8% -88.6%Energy 10.0% -64.0%Energy Services 8.8% -71.6%Financial Services -0.5% -76.6%Health Care 4.3% -41.3%Internet -12.0% -92.9%Leisure -0.8% -68.2%Retailing 0.3% -58.3%Technology -6.5% -84.0%Telecommunications -10.7% -87.8%Transport 3.0% -63.5%Utilities 1.1% -61.8%Trading System 28.3% -49.7%
with MA filter 26.5% -28.7%
TABLE 8: SECTOR FUNDS2000-10 RETURNS, DRAWDOWNS
19
2000 would have included a technology one. We will start with an equal dollar
amount in each fund one the last trading day of 1999 and rebalance to equal dollar
amounts in the four funds at the end of each year.
It would be highly unlikely that the passive allocation using four funds, none
of which had a rate of return anywhere close to the trading system, would yield
results at all comparable to those of the trading system. In a bit we will see that is
the case. To make what should be a more interesting challenge, I will “cheat” even
further by assuming owning only the sector fund with the best return for the year
that year. In other words, our crystal ball is so good it will tell us at the start of the
year which Rydex sector fund will have the highest return, but not the amount, in
the coming year. (If it could tell us the amount, we might decide not to own it if it
was going to lose or make too little. There is a limit to how much I am willing to
“cheat” for the test.) For 2000 through 2010 the funds with the best annual returns
are Energy Services, Retailing, Banking, Electronics, Energy Services, Energy
Services again, Basic Materials, Energy Services, Biotechnology, Electronics, and
finally Leisure in 2010.
We have two passive methods to challenge the trading system, and both
have the advantage of being able to own the funds with the best periodic returns.
Table 9 on the next page, which is similar to Table 7, shows the comparisons.
20
As anticipated, passive asset allocation does not come close to the trading
system. Owning the best fund each year is comparable to the trading system. It is
interesting that in some years, the trading system makes more than even the best
individual sector fund.
I believe that Table 9 makes an incredibly strong case for the superiority of
active management in a secular bear market. Although the Rydex sector funds have
not been around long enough to test during a secular bull market, it is quite likely
that active management is superior then, particularly in regard to risk reduction.
The two rivals in Table 9 have the benefit of 20-20 hindsight, and as best the
author knows, no real person or system has can forecast nearly that accurately.
To be fair, the trading system was developed with full knowledge of what
happened in 2000-2010, so it benefited in a similar way as the passive cases. The
real issue is which method is more likely to produce superior performance in the
future. It certainly won’t be a method that requires picking the best performing
2000-10: 28.2% 6.9% 29.0% 26.7% 2000-10Max DD: -49.7% -48.9% -34.9% -28.7% Average
date of: 11/20/08 11/20/08 12/1/08 9/26/01 ExposureUlcer Ind: 12.6% 14.3% 10.0% 9.3% 69.5%
UPI: 2.08 0.34 2.70 2.58
Return Max DD Return Max DD Return Max DD Return Max DD Exposure2000 92.9% -17.6% 5.1% -18.1% 41.4% -30.3% 92.9% -17.6% 100.0%2001 50.5% -35.7% -7.0% -19.2% 3.2% -28.9% 32.2% -28.7% 83.1%
2002 48.7% -27.3% -12.6% -26.4% -2.0% -28.7% 43.3% -18.0% 70.6%2003 55.0% -11.5% 27.4% -11.9% 72.7% -20.4% 60.6% -9.8% 91.3%
2004 25.5% -15.9% 18.1% -7.2% 34.7% -12.9% 20.8% -8.5% 44.4%2005 7.0% -14.6% 13.1% -8.8% 48.5% -15.3% 8.3% -9.6% 68.3%2006 5.8% -15.7% 14.0% -10.8% 22.1% -16.8% 10.2% -8.5% 57.8%
2007 22.8% -10.9% 21.4% -11.3% 37.4% -14.3% 11.1% -12.4% 74.5%2008 -21.5% -49.7% -34.7% -48.9% -10.5% -34.9% -10.0% -13.5% 37.5%2009 27.8% -33.4% 35.1% -20.7% 70.7% -19.1% 41.9% -13.8% 63.9%
2010 31.0% -13.8% 17.7% -14.3% 30.4% -19.4% 4.9% -21.8% 73.4%
TABLE 9: PERFORMANCE OF TRADING SYSTEM, PASSIVE "RIVALS", & WITH 20/100 MA FILTERTrading System Passive Allocation Best yearly fund System & MA filter
21
sector fund for a year or the four best over a prolonged period. There is no doubt
that the active management such as the trading system presented here is by far and
away the better choice. The system is simple with just four parameters and has
been developed using data from a wide variety of market conditions. While its real
time performance is almost certain not to be as good as in the backtesting, it can be
expected to produce quite satisfactory trading results, particularly in comparison to
passive methods.
Additional Considerations
1. How do we know the trading system will work in real time?
The acid test of any trading or investing method is real time, out-of-sample
experience. Since data through the end of 2010 were used for the testing and
development, not enough time has passed yet to make any meaningful evaluation.
(For the first two months of 2011, the moving average crossover has been positive
and the system has been traded. It is up 5.3%, which trails the index fund’s gain of
5.9%, but is ahead of the gain of 4.9% for the daily average sector fund/random
fund choice.)
However, the author has a considerable amount of experience with a similar
trading method. A more complex method of trading the Rydex sector fund with the
lowest CCI for a short holding period has been used for about six years. Its
unfiltered trading rules are more complex than those of the basic system in this
paper, and the filter method is based on drawdowns in the equity curve rather than
a moving average crossover. So far, the author is quite pleased with the results.
22
Wanting to find a simpler system that could be written about in a reasonable length
was one of the factors motivating the research presented here.
2. Wouldn’t walk-forward testing handle the lack of out-of-sample experience?
Both of the authors of the two books mentioned above, Howard Bandy and
Robert Pardo, are strong advocates of walk-forward testing and development being
the best way to find effective trading systems. There is much merit in what they
say, but things are not all that cut and dried. A thorough discussion of that issue is
beyond the scope of this paper. However, I will mention a couple of the reasons I
did not try walk-forward for the current system.
Although walk-forward has some aspects of testing on out-of-sample data, it
is not in real time. The reason the distinction is important is that if the walk-
forward testing leads to an unsatisfactory system due to poor out-of-sample results,
the system will be discarded and one that is at least somewhat different will be
examined. In other words, it is not truly out-of-sample because the entire data set
eventually is used to determine whether or not to give the trading method further
consideration or to discard it.
A second factor is that walk-forward adds two parameters to the system, so
it becomes more complex. If the four parameter system type described above were
developed using walk-forward testing, one would need to establish the look back
period used to determine the parameters at each phase and the length of the walk-
forward test period. For example, the parameters might be based on a one year
period and tested over the next six months. That means there are really six
23
parameters involved. If a test run was not satisfactory, the developer might decide
to vary the look back and testing periods, the two additional parameters, before
deciding the basic trading model formulation was not viable. As trading system
developers well know, the more complex a system is the less likely its real time
performance will come close to being as desirable as the backtested results.
3. Did you try to optimize all four parameters as a group?
I considered doing so, but decided not to. In effect, I optimized in two
stages. The first was to find the CCI look back length and the minimum number of
days to hold a fund. The second was to find the filter based on a moving average
crossover of the equity curve resulting from the first two parameters. Optimizing
all four parameters as a group using UPI as the measure of merit could not yield a
lower UPI and likely would result in a higher one. My thinking is that I wanted to
find an attractive trading system that could stand on its own. The unfiltered one fits
that description if one is quite risk-tolerant.
If all four parameters were optimized as a group, the basic two parameter
system quite likely would not be as good as the one I developed by itself. If the
four parameter optimization resulted in the same first two parameters (16, 3), then
the two moving average parameters would also be the same as those I developed
(20, 100). If the first two are different and the UPI is higher, then the four
parameter optimization is adjusting the moving average crossover to a less
desirable basic system. That likely decreases the chances of the trading method
working well in real time.
24
4. Can the methods work on different groups of funds?
In other words, is there something special about (Rydex) sector funds that
makes the lowest CCI method effective? One way to attempt to answer the
question is to try the approach on a different type of funds. I decided to select a
group of 16 single country sector funds and use the “technology” already
developed to see if an effective trading system would result.
There are enough iShares MSCI single country ETFs that date back to 1996
to do the testing. Unlike traditional mutual funds such as the Rydex sector funds,
trying to trade ETFs using the methods described above has some practical
difficulties. The primary one is that one can’t exchange directly from one ETF to a
different ETF. A sell order for the fund being exited and buy order for the fund
being entered are needed. This likely could not be done “at the close” because the
execution prices would not be known, and hence how many shares to buy. Two
transactions, the sell followed immediately by the buy, just before the close of
trading could come fairly close to a “perfect” exchange. Since this is a theoretical
exercise, I will assume we can exchange from one ETF to another at the closing
prices. Because there will commissions for the transactions, I did the analysis with
0.04%5 deducted for each round trip trade.
The 16 country funds are Australia, Canada, Sweden, Germany, Hong Kong,
Italy, Japan, Switzerland, Malaysia, Netherlands, Austria, Spain, France,
Singapore, U.K., and Mexico. For the single fund buy and hope rival, I chose the
5 Based on a $50,000 account and $10 per transaction. These are reasonable, but obviously there can be significantdifferences. The value of the commission charge can affect the parameters found.
25
Vanguard Total International Index fund (VGTSX). The other passive rival is
theoretical based on the average daily percent change of the sixteen country funds,
which is the same idea used with the sector funds. Cash when out of the market
would be in the Vanguard Prime Money Market fund.
Working as above, I found parameters for the basic trading system of 10
days for the CCI moving average period and a minimum of 5 days to hold the fund
purchased. The moving average crossover parameters for the filter are 15 days and
35 days. These values are quite a bit different from those for trading the sector
funds. However, that is neither surprising nor a significant concern. There is little
reason to think there is a “universal” set that works well for many classes of assets.
As above, the measure of merit is the ulcer performance index for the period 2000
through 2010. Table 10, the equivalent of Table 7, shows how the trading system,
basic and filtered, and the two buy and hold rivals fared.
2000-10: 16.8% 3.0% 6.1% 21.7% 2000-10Max DD: -56.3% -61.5% -58.3% -31.6% Average
dateof: 3/9/09 3/9/09 3/9/09 8/20/10 ExposureUlcer Ind: 19.1% 26.5% 20.5% 11.2% 64.5%
UPI: 0.73 0.01 0.16 1.68
Return MaxDD Return MaxDD Return Max DD Return Max DD Exposure2000 36.2% -24.0% -15.6% -20.1% -10.5% -16.0% 34.0% -17.7% 71.8%2001 -1.9% -29.1% -20.2% -32.6% -14.3% -32.1% 6.1% -12.4% 51.6%
2002 -8.4% -33.1% -15.1% -27.6% -10.3% -26.8% 12.6% -11.1% 43.7%2003 21.7% -13.6% 40.3% -15.6% 43.2% -13.1% 24.9% -12.8% 80.2%
2004 15.7% -15.5% 20.6% -10.6% 24.1% -10.8% 10.2% -12.5% 63.1%2005 27.2% -11.8% 15.6% -7.1% 11.6% -6.6% 21.6% -8.4% 70.6%2006 47.5% -15.3% 26.6% -16.6% 30.9% -14.7% 51.8% -9.8% 87.3%
2007 21.2% -16.4% 15.5% -12.8% 16.8% -12.2% 5.7% -11.8% 72.5%2008 -38.9% -54.9% -44.1% -55.9% -40.7% -53.4% 4.2% -15.3% 32.0%2009 84.9% -30.7% 36.7% -27.7% 38.7% -26.9% 107.0% -21.1% 79.4%
2010 24.2% -21.1% 11.1% -18.6% 11.8% -19.1% -6.5% -31.6% 65.1%
TABLE 10: TRADING SYSTEMon INTERNATIONAL ETFs, "RIVALS", &WITH15/35 MAFILTERTradingSystem VGTSX Avg. country ETF System &MA filter
26
The overall pattern of the comparisons is the same as for the sector funds
with some significant differences. Looking at the first and last group of columns,
we see that the moving average filter both greatly reduces the risk levels of the
basic system, as was the case for sector funds, and also increases the return quite a
bit, the opposite of what happened before. The passive buy and hold alternatives
did better than those for the sector funds, but still proved much less worthy than
either version of the active management method.
Since the trading method would be a hassle to implement using the single
country ETFs and I have never attempted to do it, I won’t go into more detail or
explore the other passive approaches shown for the sector funds. The information
shown for the single country ETFs is yet another compelling illustration of the
value of active management.
5. With the three day holding period, aren’t there three streams of investment
performance that need to be taken into account?
In other words, one could have started by purchasing a Rydex sector fund on
any of the last three market days of 1998 in order to be using the system for all of
1999. However, the Energy fund had the lowest CCI for five consecutive days
twice in January 1999, so by the end of that month all three of the streams starting
in late 1998 would have owned the same fund and continued to own the same fund
afterwards. There have been numerous times when a Rydex sector fund had the
lowest CCI for five or more days in a row. The analysis and testing to develop the
27
trading system start with 2000 performance. By starting the equity curve
calculations at the beginning of 1999, there is only one stream of performance
involved in 2000 and later. The key is that the three days is a minimum holding
period, so some of the trades will last longer, which enables the eventual
confluence of the streams.
6. What percentage of the trades are profitable and what are the average returns
of the winning and losing trades?
This information is less important than what is shown in the tables above,
but it does provide more of the “flavor” of the trading system. For 2000-10 there
were 559 sector fund trades in the system with the moving average crossover filter,
which does not count the money market ones. Of those, 58.7% showed a profit.
The winning trades had an average return of 2.38%, the losing ones averaged
-2.24%, and the overall average return per non-money market trade was 0.52%.
Table 11 repeats the last three columns of Table 9 and shows the number of trades,
the number that gained, the percentage that gained, and the average trade result for
each year excluding the money market fund trades.
Return Max DD Exposure # of trades # up % up Avg. gain
2000 92.9% -17.6% 100.0% 70 42 60% 1.26%2001 32.2% -28.7% 83.1% 56 37 66% 0.68%
2002 43.3% -18.0% 70.6% 58 33 57% 0.60%2003 60.6% -9.8% 91.3% 68 43 63% 0.68%2004 20.8% -8.5% 44.4% 33 23 70% 0.49%
2005 8.3% -9.6% 68.3% 53 29 55% 0.32%2006 10.2% -8.5% 57.8% 42 23 55% 0.10%2007 11.1% -12.4% 74.5% 53 27 51% 0.10%
2008 -10.0% -13.5% 37.5% 29 15 52% -0.20%2009 41.9% -13.8% 63.9% 46 27 59% 0.82%
2010 4.9% -21.8% 73.4% 51 29 57% 0.12%2000-10 26.7% -28.7% 69.5% 559 328 59% 0.52%
TABLE 11: TRADING SYSTEM WITH MOVING AVERAGE FILTER
Money market trades included Money market trades NOT included
28
Summary
The relatively simple trading systems shown above for the Rydex sector
funds and also for a group of single country ETFs show the value of active
management. The systems do far better with regard to investment returns and risk
levels than the Vanguard broad market index funds that would be typical in buy
and hold accounts. Even with being able to know the best funds in advance,
passive asset allocation among the sector funds with rebalancing does not come
close to the trading systems’ performance. It takes owning the best sector fund
each year, and if that could be done it might be considered to be active
management, to get comparable returns and risk measures. So a well designed
trading system can be a practical substitute to a highly accurate crystal ball, none
of which is known to exist.
No claim is made that the lowest CCI method shown above is better than
other methods of technical analysis. I would not be surprised if someone else found
an even better relatively simple model for trading the Rydex sector funds. Such a
model would be yet another demonstration of the value of active management.
We are in the midst of a secular bear market and have seen stocks fall by
huge amounts twice since 2000. With that in mind, it is an unsolved “mystery” to
me why so many so-called “experts” still say that active management, which they
often describe as “market timing,” can’t be done successfully, so some version of
passive asset allocation is the only sensible way to achieve one’s financial goals.
29
They say that in spite of abundant evidence that passive methods do not work.
Fortunately, NAAIM members know better.