Introduction
If someone told you that they would sell you a trading system that would vastly outperform professional, institutional money managers year after year, that gives you a robust menu of expectancy vs. volatility choices, is extremely commission efficient, works in all market types, and only takes 15 minutes of work per month, how would you react?
Would you turn and walk away? Would you laugh and consider the idea preposterous? Would you be interested if you could put aside your beliefs that it is not possible or too good to be true? Well get your checkbook out! This system is only four easy payments of $1999. Just kidding.
If you did walk away, it would be the equivalent of walking past a field of diamonds. The system described above is simply a momentum-based monthly rebalancing system using ETFs. It is the simplest trading system in the world (with the possible exception to buy-and-hold and the Browne Permanent portfolio). It literally provides passive income and lives up the all the promises above. The rest of this research document provides the data that will show how this type of system lives up to the lofty claims above.
Disclaimer
This presentation is for educational purposes only and should not be construed as investment or
trading advice. Trading and investing involve significant risk of loss. You are responsible for your own
trading and investing results, and individual results vary. You must assess the risk of any trade with
your broker or financial professional and make your own independent decisions regarding any trades
or investments.
The information contained here is believed to be accurate, but no claims or warranties, implied or
otherwise, are made pertaining to the accuracy of this information and we are not responsible for
errors or omissions. Posted past results are hypothetical trades. Results do not reflect slippage,
commissions and other market risk factors inherent in actual trades in the stock market. Past results
are not indicative of future performance.
David Walton and anyone affiliated with this educational presentation assume no responsibilities for
your trading results. David Walton and anyone affiliated with this educational presentation, including
friends and relatives, may or may not take a position or have a position in these stocks, based on
their trading discretion at the time. You should assume that these positions may be substantial
enough to be material.
Acknowledgements
This research is possible thanks to the great system development work of Dr. Ken Long. Additionally, I stand on the shoulders of the great research and trading reports from Jim Carroll, Bill Scheidt, Bob Steinmetz, and many others in the Tortoise mastermind.
Research Objectives
Many talented people have tested and traded some form of monthly rebalancing system, including the concept developer Dr. Ken Long. All of the research and feedback from live trading indicates that monthly rebalancing among ETFs representing different asset classes is a robust trading concept. The research conducted for this writing is an attempt to thoroughly test the robustness of a general trading strategy through back-testing.
This research report attempts to answer two simple questions:
1) Does the monthly rebalancing strategy have a measurable edge?
2) Which combinations of parameters are particularly robust?
Limitations and Assumptions
All of the research conducted for this report was conducted in Microsoft Excel using the XLQ plug-in. The market data source used was the Yahoo adjusted close price data series. Although the data was spot-checked against other data providers, there could be some minor inconsistencies that might influence back-test results.
Further, the back-test data does not include commissions or slippage. Although a monthly rebalancing system is inherently commission efficient since it only trades one round trip per month, there is a definite performance degradation that is dependent on account size. Slippage is also a consideration and the very real effects should be determined through actual trading.
All of the results presented here are from back-test data. The results should be viewed as best case system performance and should be verified through a robust forward test prior to trading any system at full-size.
Finally, the back-test results assume that all trading decisions and order executions are conducted upon market close of the last trading day of each month. This may not be possible with all brokers. Additionally, because decisions cannot actually be made and executed simultaneously in reality upon market close, mistakes in actual trading are likely.
Methods
A total of 247 monthly rebalancing systems were back-tested using excel and XLQ (data from yahoo). These systems were comprised varying the following parameters: 1) ETF portfolio composition, 2) number of holdings per month, 3) using a 4-month MA as a stop and entry filter while using cash, SHY, and TLT when a stop or filter is active, 4) using a 3-month relative strength vs. a blended 3-month and 6-month relative strength, and 5) using leverage (in a margin account). Additionally the Browne permanent portfolio and buy-and-hold of the S&P500 systems were included for comparison.
Parameter #1: Portfolio Composition
The table below shows the ETF sets that were analyzed.
Ken's 26 / Ken's 23 / Ken's 14 / HG 23 / LR 15 / 431MR 11 / Corel 11 / Browne 4 / Combo 33 / Combo 41DBA / DBA / DBA / DBA / DBA
DBC / DBC / DBC / DBC / DBC
DIA / DIA / DIA / DIA / DIA / DIA
EEB / EEB / EEB / EEB
EEM / EEM / EEM / EEM / EEM / EEM / EEM
EFA / EFA / EFA / EFA / EFA
EMB / EMB / EMB / EMB
EPP / EPP / EPP / EPP / EPP / EPP / EPP / EPP
EWA / EWA / EWA / EWA / EWA / EWA
EWJ / EWJ / EWJ / EWJ / EWJ / EWJ
EWZ / EWZ / EWZ / EWZ / EWZ
FXF / FXF / FXF / FXF
FXI / FXI / FXI / FXI / FXI
GLD / GLD / GLD / GLD / GLD / GLD / GLD / GLD
GUR / GUR / GUR
IEV / IEV / IEV / IEV / IEV
ILF / ILF / ILF / ILF / ILF / ILF / ILF
IWM / IWM / IWM / IWM / IWM / IWM / IWM / IWM
IYR / IYR / IYR / IYR / IYR / IYR
MDY / MDY / MDY / MDY / MDY / MDY / MDY / MDY / MDY
QQQ / QQQ / QQQ / QQQ / QQQ / QQQ
SHY / SHY / SHY / SHY / SHY / SHY / SHY / SHY / SHY / SHY
SPY / SPY / SPY / SPY / SPY / SPY / SPY / SPY / SPY / SPY
TLT / TLT / TLT / TLT / TLT / TLT / TLT / TLT / TLT / TLT
XLE / XLE / XLE / XLE / XLE
XLF / XLF / XLF / XLF
SMH / SMH / SMH
XME / XME / XME
HYG / HYG / HYG
LQD / LQD / LQD / LQD
VGK / VGK / VGK / VGK
VNQ / VNQ / VNQ
XLU / XLU / XLU / XLU
SSO / SSO
EFZ / EFZ
ERX / ERX
PSQ / PSQ
TMF / TMF
TNA / TNA
TQQQ / TQQQ
USO / USO
These sets of ETFs were chosen based on previously tested systems by Dr. Ken Long, systems tested and others traded by participants of the Tortoise chat room, ETF sets specifically chosen by the author for other reasons to be mentioned, and combinations of all the above.
The author created two portfolios of ETFs named 431MR 11 and Corel 11 respectively. In both cases, combinations of 11 ETFs were chosen due to a belief held by the author that a smaller number of ETF included in the monthly selection process generates higher SQN systems. This hypothesis was tested as part of the research.
The composition of the 431MR 11 portfolio was based on several factors. The equity ETFs were chosen to allow for capturing persistent edges of seasonality among small vs. large cap and growth vs. value stocks. TLT, GLD, IYR, and SHY were chosen as non-correlated asset classes, both to each other and equities.
The composition of the Corel 11 portfolio was primarily based on picking the 5 least correlated ETFs from the larger universe of the Ken Long 26 ETF BMR system combined with equity ETFs representative of areas where persistent edges have been documented in scholarly research as mentioned above.
Parameter #2: Number of Holdings per Month
In general systems were defined to hold the top one to six ETFs based on the relative strength time period (variable #5) from the respective portfolio for one month. In the case of the Browne 4 ETF portfolio, only systems using the top one to three ETFs were chosen. Thus for a given ETF portfolio, system #1 would buy and hold only the top ETF based on relative strength. System #2 would buy and hold the top two ETFs based on relative strength, etc. The S&P 500 buy and hold holding period was for the duration of the back-test period. The Browne permanent portfolio rebalancing period was 1-year unless any component became > 35% or < 15% of the portfolio per rules described by Harry Browne.
Parameter #3: Moving Average Stop/Buy Filter and Cash Equivalent ETF Selection
In order to test the efficacy of using a long term moving average as a stop and buy filter, systems were defined to either have no stop and no buy filter or to use the 4-month moving average as both a stop and buy filter. The meaning of buy filter is as follows: if one of the top ETFs that would otherwise be selected for holding for the month is below its own 4-month moving average, the cash equivalent ETF would be held in lieu of the chosen ETF.
In this analysis, if a held ETF traded at or below the 4-month moving average within the month, the ETF would be sold and funds would stay in cash until the next month began. In other words, only cash was used for the stop loss, not a cash equivalent ETF. This was done for the sake of simplicity.
This analysis used three options for cash for when the buy filter was active. The first option was simply cash. The second option was the SHY ETF which represents short term (1-3 year) US treasury bonds. The third option was the TLT ETF which represents long term (20+ year) US treasury bonds.
Thus a total of four stop/buy filter options were analyzed: 1) 4-month MA stop and buy filter to cash, 2) 4-month MA stop to cash and buy filter to SHY, 3) 4-month MA stop to cash and buy filter to SHY, and 4) no stop/buy filter of any kind.
Parameter #4: Relative Strength Time Period
Two different relative strength measures were used to construct systems: 1) 3-month relative strength, and 2) blended 3-month and 6-month relative strength. The specific blend weightings were 70% 3-month and 30% 6-month. Both of these periods have been shown to provide a persistent edge in scholarly research.
Parameter #5: Use of Leverage
Finally, a subset of systems constructed from the variables above was chosen to evaluate the effect of leverage. In this case, the top-3 and top-4 systems using the 4-month moving average stop/buy filter and SHY as the cash equivalent were chosen. The reason for choosing those combinations is that that combination of variables was shown to be particularly robust.
Leverage was added at a level of 2x, which is achievable in a portfolio margined account without much risk of a margin call. A margin interest rate of 2% was used. This rate is just above the current margin rate at Interactive Brokers (currently 1.65% at the time of writing). It is important to note that most retail brokers charge much more than 2% in margin interest, with most charging around 8% as of this writing. Nevertheless Interactive Brokers provides both a competitive margin interest rate and portfolio margin accounts. The intent of this study is to reasonably approximate what is possible for an actual system.
Trading System Construction
Setups/Filters:
· Each month the ETFs in the population were ranked 1-n by relative strength time period (variable #5).
· If a stop/buy filter is selected, the 4 month MA of each top ranked ETF was used to determine if the top performing ETFs can be purchased. If their current price is below the 4 month MA, then the cash ETF will be used instead for that month.
· If a stop/buy filter is selected, the 4-month MA stop was applied to each ETF based on its own MA and price.
Entry Rules:
Trading decision time was upon market close on the last trading day of the month.
Long Rules:
• Always invested, in either the top 4 rated ETFs or in the cash ETF (SHY).
• Buy the top 4 in equal dollar amounts and hold for a month.
Every last day of the month on market close, the top n ETFs were bought and the currently held ETFs sold. If the system uses a stop/buy filter and any of the top n ETFs is below its 4-month moving average trading decision time, the cash equivalent ETF was purchased instead of the applicable ETF. This decision applies to each ETF individually meaning that anywhere between 0-n of the top ETFs could be purchased depending on their current price in relation to the 4-month MA.
Exit Rules:
Normal Exit:
The normal exit of this system is upon market close of the last trading day of the month when an ETF held is no longer one of the top n ETFs ranked by applicable relative strength. Regardless of how many months a given ETF is held, a “closed trade” in this system is considered to be the total return for the month from all n held ETFs. Thus and ETF may be held many months when it is outperforming but a “closed trade” will count each ETF’s contribution to total return for the month.
Capital Preservation Stop:
In the case when a stop/buy filter is used, the 4month MA is used as a stop loss point; if the stop was hit, the ETF was sold and held in cash until the recalculation at the next period.
Position Sizing Rules:
Position sizing uses equal allocation on a percent of portfolio basis. This means each held ETF used (1/n)% of available capital upon market close on the last trading day of the month.
Data Source
In order to conduct this research, a model was constructed in Microsoft Excel using the XLQ plug-in. The data source used by XLQ was from Yahoo adjusted price data. Adjusted data from Yahoo includes dividends and makes adjustments to the price series to remove discontinuities if/when an ETF trades ex-dividend. This analysis used only monthly data. Thus a 4-month moving average was constructed as the simple arithmetic average of the closes for the previous four months. Price information from a current (not completed) month was not considered in any calculation.