The DMS system trades ETFs (although can be expanded to other securities). Currently it follows 45 ETFs, ranking each of them according to a proprietary momentum formula.

Bull and bear , symbolic beasts of market trend

Highest-ranked-momentum ETFs can be considered for long trades. Lowest-ranked ETFs should be avoided or considered for short trades.

This article presents the rationale for momentum trading and two systems.  It also provides ten-years of back-test results. With user-selection options, 24 different outcomes (backtests) are presented. The worst of these outperforms a buy and hold S&P 500 strategy by a factor of three. The best of these do so by more than 20 times.

What is Momentum Investing?

A general definition of momentum and how it pertains to investing is provided by AQR:

Momentum is the phenomenon that securities which have performed well relative to peers (winners) on average continue to outperform, and securities that have performed relatively poorly (losers) tend to continue to underperform.

Efficient Market Hypothesis (EMH) advocates claim there is no mechanical trading strategy that outperforms markets on a consistent basis. Research on market efficiency began almost sixty years ago as a joint project between Merrill Lynch and the University of Chicago. Eugene Fama, considered the strongest advocate of EMH, was awarded a Nobel Prize recently for his work in this area. (To learn more about efficient markets, try this article.)

The one anomaly that EMH believers have difficulty explaining is the positive bias of momentum trading:

“Momentum, in my view, is the biggest embarrassment for efficient markets,” Fama said, admitting that he was “hoping it goes away.” [Source]

[As an aside, I had Dr. Fama as a professor at the University of Chicago (and also played basketball with him in lunch-time pick-up games). He was the second-best professor I had, second only to Milton Friedman. He was objective, intellectually honest and provided clarity in class. He would not get a Nobel Prize for his basketball ability, nor would any of the rest of us who partook.]

How is Momentum Measured?

There is no single measurement of momentum. Many technical indicators purport to measure it. Some even include momentum in their name.

Oscillators (stochastics, MACD, RSI, etc.) measure momentum for an individual stock. However, these only tell whether that stock’s momentum is accelerating or decelerating versus its own recent performance. Momentum investing requires relative measurements to determine which assets have better momentum versus others.

Measures as simplistic as RSI values for stocks can work. The decision rule would be to buy (sell) the stock with the higher (lower) value.

Relative momentum measures need not be simple, but they are required.

A Simple Momentum-based System

Some approaches to momentum are simple but may still be effective. For example, a popular subscription-based website has a model that allows users to tailor parameters to create their own momentum-based systems. This approach consists of three variables — two measures of return and one measure of volatility. Users specify the time periods over which the measures are calculated. They then weight the importance of these variables.

The system is a black box but it appears that each stock or ETF is ranked by its rank-order for the defined measures. The weightings of these variables are then applied to these rankings to create a composite value. This composite is then re-ranked. The highest (lowest) ranked assets are then bought (sold).

The system is simple and polished. As a former subscriber, I recommend it for those who want to try momentum trading. The site used to allow limited free testing and perhaps still does.

Dynamic Momentum

A dynamic momentum system is dynamic in the sense that it adjusts with market conditions. While it may begin with fixed weights or relationships as explained in the simple momentum-based system, it has the capability to self-adjust. These adjustments can be complex, affecting elements according to class groupings or in various disproportionate ways. Alterations may not even go in the same direction for all assets.

Some general examples that could be built into dynamic systems are the following:

• If general markets are strong, more weight could be applied to growth ETFs and less to fixed income ETFs.
• Higher VIX readings signal higher market risk. Different ETF classes should be treated more or less favorably under such conditions.
• Individual ETF performance may have weightings change as a result of individual ETF variability.

There examples show ways that dynamism or adaptivity can be added to straightforward momentum measurements.

The DMS System(s)

The DMS (Dynamic Momentum System) differs from other momentum approaches in that it allows variables and weights to self-modify in response to market conditions.

Three variables are user selected:

1. Trading Style
2. Number of Trades
3. Type of Trading

Three trading styles (Normal, Aggressive or Conservative) are available. These choices reflect the amount of risk that an investor wants to assume. The number of assets (in this case ETFs) to be traded is determined by the user. Finally, the user has the option to trade Long or to engage in both Long and Short trades. Users can alter any of these at any time.

The implications of various choices can be explored by backtesting. The backtests represent 24 combinations of these choices. Backtests are produced for each trading style, for each of the number of trades and for both Long Only and Long and Short Trades.

The Complexity of the DMS System

The complexity of the system is internalized. Users need not deal with its intricacies. Some of the dynamic features embedded include the following:

1. General market performance modifies weights. High VIX readings affect higher risk ETFs more than lower risk ETFs (a non-linear adjustment).
2. The volatility of individual ETFs affects the way that they are weighted. Volatility above or below normal levels changes weightings.
3. General market performance is reflected disproportionately. For example, high VIX penalizes some ETFs while rewarding others (e.g. safer ETFs like bonds).

The interactions between the weightings and the rewarding/penalizing of final rankings cannot easily be described. There are many permutations and combinations. Some are linear, some affect weightings disproportionately and some merely reward or penalize rankings. These dynamic adjustments are both complex and proprietary.

II. DMS LONG ONLY

The table below shows the results from January 1, 2007 through mid-day December 1, 2016. These represent only long trades. The style and pos columns on the right side of the table indicate which combination of style and number of positions are being traded.  For Style: 1 = Normal, 2 = Aggressive and 3 = Conservative. Pos are the number of trades per month.

The column meanings are:

• CAR is compounded annual return,
• Big Neg Trade and Big Sys Loss are adverse conditions. The first represents the worst percentage drop from any given trade. The second is more important as it represents the largest drawdown (peak equity to trough equity). Both are calculated only at ends of months.
• Car%/BigSys is compound annual return divided by maximum system drawdown. The higher this number, the better in the sense that annual returns are larger than maximum drawdown.
• The Sharpe Ratio is a common measurement used to evaluate performance.

The other columns should be obvious.

All comparisons begin with an initial equity of $10,000. Ending equity in each case would be  $10,000 higher than the profit figures shown below.

		DMS Long ONLY
		Big Neg	Big Sys	Car%/	Sharpe	# of
Profit	CAR%	Trade	Loss	Big Sys	Ratio	Trades	Win%	Style	Pos
$115,658	29.1	(16.3)	(21.7)	1.3	0.9	99	59.6	1	1
$101,847	27.6	(16.3)	(21.7)	1.3	0.9	99	58.6	2	1
$73,486	23.9	(16.3)	(20.0)	1.2	0.8	186	60.8	2	2
$69,460	23.3	(16.3)	(20.0)	1.2	0.8	186	60.2	1	2
$69,501	23.3	(16.3)	(20.0)	1.2	0.8	186	60.8	3	2
$76,460	24.3	(16.3)	(21.7)	1.1	0.8	100	59.0	3	1
$29,701	14.9	(18.3)	(24.5)	0.6	0.6	275	58.2	3	3
$30,180	15.1	(18.3)	(27.6)	0.6	0.6	276	58.0	1	3
$29,379	14.8	(18.3)	(28.0)	0.5	0.6	275	57.1	2	3
$15,997	10.1	(32.9)	(33.5)	0.3	0.4	370	55.7	2	4
$15,575	9.9	(32.9)	(33.5)	0.3	0.4	372	56.5	3	4
$15,887	10.1	(32.9)	(33.5)	0.3	0.4	372	56.2	1	4

Observations on Results

Some comments:

1. The fewer the number of positions traded, the larger the profit. That is a positive indication of the effectiveness of this momentum ranking algorithm. As more trades are taken, lower ranked ETFs are added to the mix, diminishing the profits.
2. One or two positions produce the best results.
3. Normal and Aggressive trading styles tend to outperform Conservative.
4. For those trading the Long Only system, there is no advantage to trading more than two positions, regardless of the choice of your trading style. Trading more than two positions penalizes backtest profits. 
5. Trading more positions does not reduce risk, at least as measured by Big Sys Loss (maximum system drawdown).

Diversification Comments

This trading system is risky as evidenced by the maximum system drawdowns. At best, drawdowns  of 20% or higher should be expected. Trading more positions does not provide diversification benefits. It lowers returns and appears to elevate drawdowns.

Nothing can be done using this trading system to achieve diversification. (The SSS switch that moves into and out of the system markets can mitigate risk. Its effects are not shown above.).

Anyone who trades this system should invest the bulk of his/her funds conventionally. That is, asset diversification and the beneficial effects of risk reduction must be achieved via other holdings not highly correlated with this type of momentum investing. Stated differently, funds committed to the DMS Long Only strategy should be a small part of one’s overall investment program.

III. SIMPLE SWITCHING SYSTEM (SSS)

Earlier explanations of the DMS system used a Simple Switching System (SSS) as a modifier to the DMS results. The SSS switches between cash and the DMS Long Only system. It serves to reduce returns and risks. A review of earlier articles under the menu choice Investing will find these articles.

For purposes of this article, SSS is ignored so that the DMS Long and Short System may be explored. The SSS will be provided as a trading tool, it just will not be dealt with in this overly-long article. A future article is likely on this topic.

IV. DMS LONG AND SHORT MOMENTUM SYSTEM

Momentum can be used to trade long or short. The higher the momentum the more likely a stock or ETF will outperform the market. The lower the momentum, the more likely it will underperform. Short selling is a way to capitalize on the expected underperformers.

Caveat

The ability to sell short (or at least the ability to see the effects of doing so) has been added to this model. That is not a recommendation to engage in short-selling which is riskier than long-only trading. Higher risk comes from the fact that short-selling affords the possibility of losing more than you invested.

Short-selling is prohibited in IRA accounts. To sell short one must use a regular brokerage account or utilize contra ETFs to approximate the same results within an IRA. For major ETF categories like the indices or bonds, contra accounts are available. More narrowly defined ETFs may not afford this opportunity as easily.

Again, for backtesting purposes, trading preferences and number of trades is held constant across the backtest period. Also, the actual ETFs are traded as if they could be shorted. In reality, the supply of securities available for short-selling may restrict one from accomplishing short sales even if he/she is trading a regular brokerage account. 

Inverse ETFs are under review. While they will not be included in the momentum calculations, they may help in getting around this problem. If there are inverse funds that match up to selections when a low-ranked ETF is identified as a potential short, the inverse fund may be suggested. There is no assurance that enough inverse funds will meet the requirement (liquid 1x inverse funds nearly 100% negatively correlated to funds in the working list of 45 ETFs).

New System

The DMS Long and Short system is identical to the DMS Long Only system in terms of user choice. What is different is that for more than one trade per month, the results include short trades. If two ETFs are traded, one is traded long and the other short. Long takes priority, so if you trade only one ETF, it will be a long position. If you trade two it will be one long and one short. Three is two longs and one short and four is two of each.

Backtest results for the long and short strategy are shown below. Columns and time periods are the same as in the prior table.

		DMS Long & Short
		Big Neg	Big Sys	Car%/	Sharpe	# of
Profit	CAR%	Trade	Loss	Big Sys	Ratio	Trades	Win%	Style	Pos
$115,658	29.1	(16.3)	(21.7)	1.3	0.9	99	59.6	1	1
$101,847	27.6	(16.3)	(21.7)	1.3	0.9	99	58.6	2	1
$76,460	24.3	(16.3)	(21.7)	1.1	0.8	100	59.0	3	1
$56,309	21.0	(32.8)	(15.5)	1.4	0.6	203	55.7	2	2
$60,324	21.8	(32.8)	(15.5)	1.4	0.7	203	56.7	3	2
$71,740	23.6	(32.8)	(15.5)	1.5	0.7	201	56.7	1	2
$61,126	21.9	(32.8)	(12.9)	1.7	0.7	288	58.0	1	3
$63,048	22.2	(32.8)	(12.9)	1.7	0.7	289	58.5	3	3
$55,966	21.0	(32.8)	(12.9)	1.6	0.7	290	57.9	2	3
$49,848	19.8	(32.8)	(11.3)	1.8	0.6	383	56.9	3	4
$47,133	19.2	(32.8)	(12.0)	1.6	0.6	383	56.7	2	4
$43,961	18.5	(32.8)	(12.0)	1.6	0.6	384	56.5	1	4

Observations on Results

Here are some observations based on the above table::

1. Styles with only one trade produce the same results as in the Long Only table because the position is always long. The rest of the results differ.
2. While more trades generally produce lower profits, the drop-off in profitability is not nearly as severe as the DMS Long Only system.
3. Diversification benefits, at least in terms of maximum system drawdown, occur with this system as more trades per month are taken.
4. Two positions (one long and one short) produces comparable profit to the DMS Long Only 2 trades. It does so with a few more trades (about 10% more) but shows lower risk as reflected in the maximum drawdowns.
5. Beyond two trades, this system is superior to the DMS Long Only system in both profit and risk.

V. S&P 500

A trading system should be evaluated against a benchmark to assess its efficacy. A common benchmark is the S&P 500 index. It is reasonably well diversified and provides an approximation of a “buy and hold” strategy. Advocates of long-term investing often use this measure as a proxy for what investors can expect to earn.

Buy and Hold

SPY, the ETF which is designed to duplicate the S&P 500, will be used in the test. SPY is a reasonable approximation of what holding the entire S&P 500 would produce.

If the momentum-based systems work properly, one would expect them to produce higher returns than a buy and hold strategy. Similarly, because they invest in fewer assets, one might expect them to be higher risk than SPY which holds 500 stocks.

The results are somewhat surprising:

• Buying SPY January 2007 with $10,000 and holding it until Dec. 1, 2016 produced a profit of $5,491. This profit is dwarfed by all of the 24 backtests shown above. The magnitude of difference is startling! The best of the momentum-based strategies produced returns twenty times better than buy and hold. The worst were triple the SPY outcome.
• Using maximum system drawdown as a proxy for risk, SPY (the S&P500) shocked again. The buy and hold strategy produced a maximum system drawdown of 55%, despite having almost 500 more stocks than the momentum systems.

The buy and hold strategy, in short, produced terrible returns and did so at extremely high risk.

Is this type of momentum trading system the answer? During this time period, an affirmative appears to a proper answer. However, these results are backtest results. There is no guarantee that such performance will continue in the future. Markets change and so do the efficacies of systems. The period from 2001 through 2016 had some rough road along the way. Will the system hold up in the future?

One should never assume that he has found the Holy Grail. Rarely can an investor both sleep well and eat well.

Another Comparison

To say that one system is better than another based on a few statistics is dangerous. Hidden in summary statistics could be very good or very poor periods. One can examine the past a bit more closely.

One way to do that is to look at individual annual returns in terms of their consistency. Is a positive (or negative) ten-year result due to one or two great (or poor) years? Does a system show consistency? Is the underperformance of SPY due to the market collapse that began in late 2007 only?

To better assess such questions, a look at annual returns can be useful. For purpose of the review, the DMS Long Only system and the DMS Long and Short System will be compared to SPY on an annual basis. For both the DMS Long and DMS Long and Short, normal trading style of two positions will be used.

The following table compares annual percentage returns of the DMS Long & Short system (on the left) and the DMS Long Only system (on the right) with SPY (which is labeled as BH Yr%). The greenish yellow boxes to the left of each column show where the momentum systems outperformed SPY in a year by 5% points or more. Likewise, the Red arrows show where SPY outperformed the momentum system by more than 5% points.

 DMS SYSTEM

 DMS LONG & SHORT   DMS Long ONLY

From the above table, the following comments:

• In both instances, SPY only outperformed its momentum comparison two years out of ten.
• Momentum systems outperformed. In one instance they did so in seven years; in the other, five years.

Based on this time frame and the algorithm employed in the momentum systems, either can be claimed superior to a buy and hold strategy using the S&P500 as a benchmark. The smaller segments (yearly results) support the original overall analysis.

Further Disaggregation

The results can be further disaggregated to continue to inspect the relative performances of each of the systems versus the benchmark or against each other. For example, each of the systems could be broken down into monthly returns for each of the years and compared to SPY. Further, each could be broken down into the trades made by month. Then each of the trades could be traced back to the relative rankings of each of the securities traded and not traded. For purposes at hand, we have already gone too far for most readers. To end, the monthly profit percentages for the two trading systems shown above is provided below. Readers are free to apply their own interpretation and analysis to these.

DMS LONG ONLY

DMS LONG AND SHORT

Part II of this series deals with rankings and the mechanics of using the system. That is, rankings are traced to buys and sells and tied back into individual gains or losses from trades.