The iM-5 ETF Trader

  • This system always holds five ETFs (equity-, fixed income-, leveraged equity-, short equity-, and Gold-ETFs) selected according to stock market climate and rank.
  • Typically, during good-equity markets it holds equity-ETFs and/or leveraged-equity ETFs, and during bad-markets fixed income-ETFs and/or short equity-ETFs. Also at times it can hold three gold-ETFs with other ETFs.
  • A one factor ranking system selects five ETFs from a preselected list of 33 ETFs. A simulation from 2000 to 2017 shows a 35% annualized return with a maximum drawdown of -13%.

The model was backtested on the on-line simulation platform Portfolio 123 which also provides extended price data for ETFs prior to their inception dates calculated from their proxies. ETFs, other than P123 extended ETFs, were only considered for selection six months after their inception. Trading costs, including slippage, were assumed as 0.1% of the trade amounts using closing prices.

Model Philosophy

The basic approach is to invest in five equity-ETFs during up-market periods and in five fixed-income-ETFs during down-market periods (Basic model). Market timing rules, listed further down, were applied to identify those periods.

To improve return and Sharpe ratio, the model invests in three gold-ETFs when gold timing rules indicate this to be profitable and permitted by the ranking system. During those periods the model could have a 60% investment in gold-ETFs.

Further improvement to annualized return is achieved by adding inverse- or leveraged equity-ETFs during down- or up-markets, respectively. The table below shows returns for all possible ETF type combinations with the Basic model. The model on top of the table has the highest return, high Sharpe Ratio with reasonable maximum drawdown and is the one to which this model description applies

width="640"/(click to enlarge)

The Ranking System

The ranking system’s approach assumes trading ETFs, rather than investing for longer periods.

The one factor system is based on the price changes over a short period. The idea being that ETFs which have experienced a decline over a short period will bounce back, reverting and doing better than ETFs which have not declined in this way.

When testing this simple ranking system on the entire universe of all ETFs and CEFs traded on US markets one finds that, when rebalanced weekly, it produces a well-defined performance decline from higher to lower ranks, as shown in the diagram below.
/

Also, one would expect that an efficient ranking system would produce higher returns for lower number of holdings in the model. This is indeed what happens when varying the number of holdings from two to ten in the ETF Trader Basic model. Trading costs were set to 0% so as not to affect the test. The model holding two ETFs produced an annualized return of 25.4%, which diminishes consistently as holdings were increased, all as shown in the table below.

width="640"/(click to enlarge)

The ETF Selection List

The model typically holds five ETFs from the list below, periodically selected by the ranking system and buy rules.

Ticker

ETF Name

AssetClass

AvgDailyTot $-million

Inception

 

IWB

iShares Russell 1000

Equity Std Long

91.1

5/19/2000

E

IWF

iShares Russell 1000 Growth

Equity Std Long

178.6

5/26/2000

 

IWV

iShares Russell 3000

Equity Std Long

19.0

5/26/2000

E

USMV

iShares Edge MSCI Min Vol USA

Equity Std Long

65.9

10/20/2011

E

MTUM

iShares Edge MSCI USA Momentum Factor

Equity Std Long

59.2

4/18/2013

 

VIG

Vanguard Dividend Appreciation

Equity Std Long

58.5

4/27/2006

E

VNQ

Vanguard REIT

Equity Std Long

417.2

9/29/2004

E

VOE

Vanguard Mid-Cap Value

Equity Std Long

27.0

8/24/2006

E

NOBL

ProShares S&P 500 Dividend Aristocrats

Equity Std Long

11.2

10/10/2013

 

PXLG

PowerShares Russell Top 200 Pure Growth

Equity Std Long

2.6

6/16/2011

 

QQQ

PowerShares QQQ Trust Series 1

Equity Std Long

2,869.2

3/10/1999

 

RSP

Guggenheim S&P 500 Equal Weight

Equity Std Long

53.4

4/30/2003

E

SPY

SPDR S&P 500 ETF Trust

Equity Std Long

14,087.1

1/29/1993

 

 

UWM

ProShares Ultra Russell2000

Equity Levgd Long

13.5

1/25/2007

E

DDM

ProShares Ultra Dow30

Equity Levgd Long

14.0

6/21/2006

 

SSO

ProShares Ultra S&P500

Equity Levgd Long

139.3

6/21/2006

E

QLD

ProShares Ultra QQQ

Equity Levgd Long

64.0

6/21/2006

 

SDS

ProShares UltraShort S&P500

Equity Levgd Short

67.1

7/13/2006

E

 

ANGL

VanEck Vectors Fallen Angel High Yield

Fixed Income

10.7

4/11/2012

 

CIU

iShares Intermediate Credit Bond

Fixed Income

40.4

1/11/2007

E

CSJ

iShares 1-3 Year Credit Bond

Fixed Income

46.0

1/11/2007

 

EMB

iShares JPMorgan USD Emerging Mkts

Fixed Income

230.5

12/19/2007

E

IEF

iShares 7-10 Year Treasury Bond

Fixed Income

192.2

7/26/2002

E

IEI

iShares 3-7 Year Treasury Bond

Fixed Income

29.8

1/11/2007

E

LQD

iShares iBoxx Investment Grade Corp

Fixed Income

355.8

7/26/2002

E

SHY

iShares 1-3 Year Treasury Bond

Fixed Income

79.2

7/26/2002

E

BND

Vanguard Total Bond Market

Fixed Income

147

4/10/2007

E

BNDX

Vanguard Total International Bond

Fixed Income

34.9

6/4/2013

E

VMBS

Vanguard Mortgage-Backed Securities

Fixed Income

25.5

11/24/2009

E

VWOB

Vanguard Emerging Markets Govt

Fixed Income

5.7

6/4/2013

E

GLD

SPDR Gold Trust

Commodities

853.5

11/18/2004

E

IAU

iShares Gold Trust

Commodities

98.3

1/28/2005

E

DBP

PowerShares DB Precious Metals

Commodities

2.1

1/5/2007

 

  E = P123 extended data from 12/31/98

 

Market timing Rules

Up- and down-markets definition is based on:

(Risk Premium= SP500 Estimated Earnings Yield – 10Y T-Note Yield)

Down-markets are defined as periods when up-market conditions are absent.

Definition of Gold-markets is based on the algorithm of the iM-Gold Timer.

 

Buy- and Sell Rules

Buy highest ranked and:

Buy Gold-ETFs when gold-buy-signal is present, or
buy Fixed Income-ETFs, or Inverse-ETFs when down-market conditions exist, or
buy Equity-ETFs or Leveraged Equity-ETFs when up-market conditions exist.

Sell-rule is independent of rank:

Sell Gold-ETFs when gold-sell-signal is present, or
sell Equity-ETFs or Leveraged Equity-ETFs when down-market condition exist, or
sell Fixed Income-ETFs, or Inverse-ETFs when up-market conditions exist.

Performance of the iM-5ETF Trader

Performance 2000-2017

Performance from Jan-2000 to May-2017 is shown in Figure-1. The model showed an annualized return 34.9% with a -13.1% maximum drawdown.

width="640"/(click to enlarge)

Performance 2009-2017

The simulated performance from Mar-9-2009 to May-2017 is shown in Figure-2. The start date for this period is the date when the S&P 500 was at its lowest level during the financial crisis recession. For the approximately 8-year backtest period the simulated annualized return was 38.1% with a maximum drawdown of -13.1%. The model significantly out-performed with lower drawdown the SPDR S&P 500 ETF (SPY) over this up-market period.

width="640"/(click to enlarge)

Performance Histogram 2000-2017

Rolling 1-year returns with a 1 week offset are shown in Figure-3. There were 3 out of 855 samples with a small negative 1-year return of about -1.8% to 0.0%, with trading costs included.

width="640"/(click to enlarge)

Calendar year performance

Calendar year returns are shown in Figure-4. There was never a year when the model had a negative return.

width="640"/(click to enlarge)

Distribution of monthly returns

Monthly return distribution is shown in Figure-5. There were only 54 negative monthly returns out of 204, versus 84 for SPY. Also, all monthly returns are within three standard deviations away from the mean, indicating that performance is not due to a few outliers with extreme returns.

width="640"/(click to enlarge)

Trading Statistics

This is a trading model with an average annual turnover of about 1,090% (11 x). The average holding period of a position was 33 days, 62% of all trades were winners, and the biggest loss of a trade was -15.6%, all as shown in the table below.

width="640"/(click to enlarge)

Risk Measurements

Risk measurements are from Portfolio 123.

width="640"/(click to enlarge)

Portfolio size and trading costs

Slippage including brokerage fees 0.1% of trade amounts was assumed in the backtest. Slippage is related to trade size, average volume traded, trade duration, daily volatility and shares outstanding. The slippage formula can be found in this article, and the calculation for the estimates of daily volatility is located here.

As portfolio size increases the slippage also increases. Since this model trades in many liquid ETFs slippage should not exceed the assumed value until the total portfolio size becomes very large.

Following the model

This model could be of interest to investors willing to accept a fair amount of trading activity. See also “Best Practices” for ETF Trading.

Note, that this is a trading model with 80% of all trades having a holding period of three weeks or less. There were 54 trades per year on average, and the maximum was 127 trades in 2015.

Holding
period

Percent
of all trades

Number of trades
2000-2017

7-21 days

80.3%

752

22-42 days

7.4%

69

43-92 days

6.6%

62

93-183 days

2.5%

23

184-240 days

0.6%

6

241-365 days

0.7%

7

More than a year

1.8%

17

On iMarketSignals we will report the performance of this model, with weekly trading signals normally provided on Sundays to Gold level members.

On 5/15/2017 the model held: DDM, QLD, SSO, UWM and VOE.

Disclaimer

Note: All performance results are hypothetical and the result of backtesting over the period 2000 to 2017. Since performance is greatly dependent on market-timing rules, the future out-of-sample performance may be significantly less if those rules are not as effective as they were during the backtest period. No claim is made about future performance.

Posted in blogs, featured, Publish
5 comments on “The iM-5 ETF Trader
  1. vman says:

    Anton and Georg

    Good one! Are the 5 ETF’s equal weighted at all times??

    Thanks
    Vman

  2. randyfloyd says:

    Hello,
    regarding ETF best practices, do you generally advocate inputing transactions for execution at open of business on monday, or making trades later on monday morning?
    thanks

    • geovrba says:

      Vanguard’s ETF best practices advises never to trade in the first and last half hour of the day.

      Our models assume trading on the first trading day of the week, and the 5ETF Trader’s simulated returns are calculated from the closing prices assuming slippage of 0.1%.

  3. nmtdoc says:

    Georg and Anton,
    Can you elaborate a bit more on the ranking system ? Am I correct in understanding that it is based on mean reversion rather than momentum?
    It would be very helpful if you would construct a correlation table of the newer models you have developed, or possibly all the models in the performance table. I imagine there are quite a few of us trading multiple models with no real understanding of how they are potentially correlated.
    Thanks,
    Jon

Leave a Reply

With reference to Section 202(a)(11)(D) of the Investment Advisers Act: We are Engineers and not Investment Advisers, read more ...
By the mere act of reading this page and navigating this site you acknowledge, agree to, and abide by the Terms of Use / Disclaimer
Share This