A Global Market Rotation Strategy With An Annual Performance of 41.4% Since 2003 - Seeking Alpha

9/7/2018 A Global Market Rotation Strategy With An Annual Performance Of 41.
4% Since 2003 | Seeking Alpha
Portfolio Strategy
A Global Market Rotation Strategy With An Annual Performance Of

41.4% Since 2003
Aug. 9, 2013 10:30 AM ET387 comments | 1 Like
by: Frank Grossmann
Sounds too good to be true! My first contribution on Seeking Alpha.
Before I get started, let me introduce myself. My name is Frank Grossmann and I am
living in Switzerland. Until recent, I used to work as a chief scientist for an international
digital imaging company. They have incorporated my software algorithms for digital image
processing, which I had developed in my company Colour-Science.com. These algorithms
perform image enhancement, red eye removal, and face or general pattern detection.
However, my passion was always to search for patterns in financial data and to develop
and back test intelligent, rule-based investment strategies.
The following strategy is one of my favorite rotation strategies, which many of my friends,
customers and I have used now for some years.
The Global Market Rotation Strategy (NYSE:GMR)
The GMR Strategy switches between 6 different ETFs on a monthly basis. The back
tested return of this strategy since 2003 is quite impressive.
Annual performance (OTCPK:CAGR) = 41.4% (S&P 500=8.4%)

Total performance since 2003 = 3740% (S&P 500=134%)
69% of the monthly trades have positive return versus 31% with negative return
https://seekingalpha.com/article/1622832-a-global-market-rotation-strategy-with-an-annual-performance-of-41_4-percent-since-2003 1/96
9/7/2018 A Global Market Rotation Strategy With An Annual Performance Of 41.4% Since 2003 | Seeking Alpha
These global markets and ETFs are:
US Market (MDY- S&P MidCap 400 SPDRs)

Europe (IEV- iShares S&P Europe 350 Index Fund
Emerging Markets (EEM- iShares MSCI Emerging Markets)
Latin America (ILF- iShares S&P Latin America)
Pacific region (EPP - iShares MSCI Pacific ex-Japan)
During market corrections I invest in:
US Treasury Bonds (EDV- Vanguard Extended Duration Tsy (25+yr))

Cash or SHY (SHY- Barclays Low Duration US Treasury)
For the design of a well performing rotation strategy, it is important that the selected ETFs
are not too volatile, show longer-term visible trends and have a good market volume, so
that they cannot be manipulated. They all should have more or less the same volatility.
The 5 global markets ETFs fulfill this condition. They all are capitalized enough, so that
they cannot be manipulated in the short term. They follow slightly different economic
cycles and have long periods where one market outperforms the others until it becomes
so overpriced and investors begin to remove their money from that market in order to
invest in other cheaper valued markets.
Looking back 12 months, we see that the US market was the clear winner and the MDY
S&P Midcaps performed +35%. In comparison, the ILF Latin America market lost nearly
-10% and the more important EEM Emerging Markets ETF only made +3.5%. Rest
assured that this outperforming will not continue forever. At one moment EEM, ILF or
another ETF will bounce back and outperform the US market. This is the moment when I
switch the ETFs.
However, switching between these markets alone is not enough to really get a good
performance. By switching only between these 5 ETFs I would have made 21% annual
return since 2003. Not too bad, compared to the 8.4% of the SPY S&P 500. But if I look at
the charts, I recognize that during periods of big market corrections also all my 5 ETFs
suffered big losses. Although these markets are globally distributed, the correlation
between them is very high. The 60-day correlation is mostly higher than 0.75. This means
there is no way to escape a market correction like the subprime crash in 2008 when the
S&P 500 lost nearly 55%. To avoid this scenario, we would need to switch to some
negatively correlated assets. The best are US Treasuries and if they do not work, then
cash is an option as well.
Based on my strategy back tests, I have chosen the EDV - Vanguard Extended Duration
Treasuries (25+yr) ETF. At the moment, due to rising yields, the correlation to the S&P500
is nearly 0 which is quite abnormal. Normally the correlation is between -0.5 and -0.75.
However, I am sure that during a future market correction these correlations will go down
very quickly.
When designing a good rotation strategy, the Return to Risk ratio (Sharpe-Ratio) is one of
the most important numbers. The higher it is, the more return you will get for the risk you
have within your investment.
Return to Risk ratio (Sharpe ratio)
Strategy (investment 2003 - today) Sharpe ratio Volatility Annual return
SPY S&P 500 investment 0.38 20.3% 8.6%
5 Global Markets alone 0.73 28% 21.3%
5 Global Markets + Cash exit 1.25 23.3% 33.8%
5 Global Markets + Treasury 1.38 25.6% 41.4%
AGG - iShares Core Total US Bond (4- 0.40 5.4% 4.3%

5yr)
As you can see, a good rotation strategy is always better than the other strategies. Many
conservative and cautious investors think that they have a small risk investing in a bond
ETF like the AGG. However, if you compare the Sharpe ratio, you recognize that an SPY
investment has about the same risk under the condition, that you invest only half your
money and keep the rest in cash. The SPY investment return would be the same,
because the annual performance was exactly twice the annual AGG performance.
With the Global Market Rotation strategy, you can reduce your risk and your volatility
easily by just keeping a good part of your money in cash or short-term bonds.
Even if the volatility of 25.6% seems a little bit high, this does not mean that the strategy is
risky. The 25.6% is the medium 20-day volatility since 2003, which is slightly higher
because of the switching between the ETFs. However, it is much more important, what
happens during a market correction. Here my strategy showed its strength in the past.
During the S&P 500 market crash in 2008, the strategy produced a solid return of 55.5%
while the S&P 500 was -36.8% down by the end of the year.
I consider such a strategy as quite a save, because the strategy will avoid big losses
during market corrections. It will always switch quite early into treasuries or into cash for
the case that treasuries would not anymore play the role of a safe haven asset.
Annual performance of the Global Market Rotation Strategy
Year GMR performance S&P 500 performance
2003 34.9% 28.2%
2004 24.6% 10.7%
2005 36.8% 4.8%
2006 39.4% 15.8%
2007 38.5% 5.1%
2008 55.4% -36.8%
2009 56.7% 26.4%
2010 40.0% 15.1%
2011 89.4% 1.9%
2012 14.2% 16%
2013 until end July 19.4% 21.2%
How to rotate the ETFs
I rotate or switch the ETFs on a monthly basis. Every month I calculate a ranking of the 6
ETFs and based on the results I invest in the best ETF for the next coming month. I use a
software, which I have written in Mathlab, to find out which the best look back period a and
the best investment period b is. The period b will give me my investment return. The
software just varies the look back period a between 1 week to 12 months for investment
periods b of 2 weeks to 3 months. The software just loops through 10 years of historical
price data and outputs a scatter diagram, which shows which a and b periods give the
best positive return. The result is a 3 dimensional mountain diagram, which highlights the
best periods. The investment period b is less critical than a. However b should always be
shorter than a. One month for b is a good value for our 5 ETFs. The look back period a
can make big differences. However, a normally lies somewhere between 1 and 5 months.
My Mathlab method is very interesting to test if an ETF can be used for a rotation or a
trend following strategy. You cannot use an ETF if you cannot find a look back-investment
period pair, which gives you a stable positive return. Also for some ETFs you get
completely different returns, only by changing the parameter a little bit. Such ETFs are
also not good for rotation strategies.
For example, it is nearly impossible to design good rotation strategies for commodities.
They are too volatile and they can be very easily manipulated. Also, if you look back a and
investment periods b are too short, it becomes difficult to achieve positive returns. The
shorter the periods are, the more your performance becomes a function of the random
market noise which overlays existing trends. This is why I am not a friend of intraday
trading. In addition intraday trading is really a hard work. I'd rather make money without
being shackled to a computer screen the whole day.
For the ranking, I also use the volatility of the ETFs. While this is not so important for the 5
Global market ETFs, it is important to lower the EDV ranking a little bit, according to the
higher volatility of the EDV ETF. EDV has a medium 20-day volatility, which is roughly 50%
higher than the volatility of the 5 global market ETFs. This results in higher spikes during
small market turbulence and the model would switch too early between shares (our 5
ETFs) and treasuries .
This probably sounds now very complicated, but if you want to play this strategy on your
own, you can get quite good results by just looking at the 3-month historical performance
and then always invest the next month in the best of the 6 ETFs. This is something you
can do easily by yourself. You will still make about 34% annual return per year. The
additional annual return I get for these complicated calculations, however, is +7.4% which
is also not too bad.
Here is an example of the investments since September 2012
Date Portfolio GMR S&P 500
month ETF monthly return monthly return
Sep.2012 EPP 2.8% 2.5%
Oct.2012 EPP 2.5% -1.8%
Nov.2012 EPP 1.8% 0.6%
Dec.2012 EPP 3.8% 0.9%
Jan.2013 EPP 3.5% 5.1%
Feb.2013 MDY 0.8% 1.3%
Mar.2013 MDY 4.8% 3.8%
Apr.2013 MDY 0.6% 1.9%
May.2013 MDY 2.3% 2.4%
Jun.2013 MDY -2.3% -1.3%
Jul.2013 MDY 6.7% 4.8%
As you can see, since January the strategy is invested in MDY. It is not necessary to
switch the ETFs every month. I would say that you have to switch all 2-3 month on
average. In addition, instead of buying the ETFs, you can buy Comex futures for some of
the ETFs. This is what I do. It is the case for MDY and IEV (Europe). Instead of the EDV
you can buy Ultra-TBonds, which are similar to the TLT ETF, but you have to buy about
1.5x your normal investment because EDV behaves like a 1.5x leveraged Ultra-TBond.
Disclosure: I am long MDY, ZIV, SLV. I wrote this article myself, and it expresses my own
opinions. I am not receiving compensation for it. I have no business relationship with any
company whose stock is mentioned in this article.
Comments (387)
varan
With 3 month look back you get 33.3% CAGR with Sharpe of 1.16 and Sortino of 2.84 and maximum drawdown of
19%, which is excellent. However, the 2012 and 2013YTD returns are not that great (about 1%).
Using TLT instead of EDV gives you a CAGR 31%, Sharpe of 1.26, Sortino 2.95 (wow!) and drawdown of 17% but
with 2012 and YTD returns of 8% and 17% respectively. Excellent.
Even without the TLT and EDV you get respectable CAGR in the 15-20% range but with high losses in 2008 and
2011. Using TLT here is what I call polygamous paired switching, a natural extension of paired switching.
I would not suggest the optimization as that might lead to overfitting.
(I buy on the first trading day of the month on the basis of the return calculated from adjusted closing prices on the
last trading days of the fourth month and first month prior to the current month. For example, on the first trading
day of May I use the adjusted closing prices on the last trading days of January and April.)
I use a script in Mathnium.
09 Aug 2013, 10:54 AM
Frank Grossmann, Contributor

Author’s reply » The Treasuries plays the role of automatic braking system of a rotation strategy. Think of
the 5 global markets be a truck speeding down the highway. Now I am following close with my car and my
car has one of these new automatic brake systems which brake automatically to avoid a crash. These
braking systems measure the distance to the car in front of you. Such a control system has a gain setting
which has to be fine tuned very well. If the gain is too low, the car will not brake enough. If the gain is to
high the car will begin to brake hard and accelerate which will cause oscillation.
You can tune the gain of the treasury by selecting the duration of the treasury. Duration is like a leverage.
If you choose a short duration Treasury it will take too long to switch and you will suffer losses at a sudden
market correction. If the duration or also leverage is too high you will switch too early which results in buy
high sell low.
Here I tried all Treasuries from SHY, IEI TLH, TLT, EDV, and leveraged Treasuries and EDV had the best
"gain" for the "automatic brake system".
10 Aug 2013, 12:18 PM
varan
Of course. TLT does wonders for tactical strategies.
10 Aug 2013, 06:19 PM
darrenstory
What about using ZROZ rather than EDV and/or TLT?
31 Oct 2013, 12:51 PM
Kevin Wilbur, Contributor

It doesn't sound too good to be true at all. Interesting inquiry. Good approach. Nice work, nicely presented and
done. Try testing it using correlated composite index or futures market proxies to your primary ETF's going back
30 years, to the beginning of the Reagan Lift.
Thanks for sharing, Frank. And welcome to the community.
09 Aug 2013, 11:14 AM

Author’s reply » I mostly calculate my strategies with ETF's but if ever possible I will trade futures. It is a
good idea to use futures data however most people rather invest in ETF's and they believe more in
strategies which are directly derived from the ETF's they invest.
10 Aug 2013, 12:18 PM
craigour
Hi Frank,
Thank you for sharing this very interesting result.
Very good job!
There are two points I am not sure if I understand correctly.

Would you kindly help?
1. In ranking, are you only using last "a" day's volatility as ranking criteria? Or some other performance measure,
like information ratio ?
2. How do you determine it's time to switch from equity to "bond"?
09 Aug 2013, 11:23 AM

Author’s reply » You can use the 20 day volatility averaged over 3 month.
For the ranking I calculate the 3 month performance of all ETF's and normalise between 0-1. The best will
have 1. Then I calculate the medium 3 month 20 day volatility and also normalize from 0-1.
Then I used Ranking= 0.7*performance +0.3*volatility.
This will give me a ranking from 0-1 from which I will take the best.
10 Aug 2013, 12:18 PM
Baboon
Performance can be negative actually, < 0.
10 Aug 2013, 01:55 PM
Ironmjolnir
Would you mind explaining why you used 0.7 and 0.3 as the values you multiplied performance and
volatility against as opposed to, say, a 60-40 or 50-50 split?
12 Aug 2013, 12:43 PM
sofocal
Hello Frank,
Thank you for the very interesting paper.
I have two questions:
1.How did you do back testing up to December 2002, if two of ETFs in your strategy have been insepted
after that: EDV- on 10/6/2007 and EEM-on 4/7/2003?
2.Is there a special reason to use equation above for introduction of volatility in your ranking (0.7 *
performance +0.3*volatility)?
Could you use directly Sharpe ratio instead which is ~ performance/ volatility?
Thank you
09 Sep 2014, 11:18 AM
thjames
did you get an answer to this? maybe he used funds or index instead?
23 Feb 2015, 12:03 AM
ramsadagopan
Excellent work.
Not having access to the exact algorithm, 3 month momentum strategy is the available option to the readers. How
does 3 month momentum strategy compare to other periods, say 1, 6 and 12months? If you are thinking about
updating this work in the future, you may want to include max drawdown, frequency of trades and Sharpe Ratio
for the sake of competeness. Also curious to see if REIT index etf would be a contender to this portfolio?
09 Aug 2013, 11:59 AM

Author’s reply » Max drawdown was 31% versus 55 for SPY in 2008 however by end of 2008 the
strategy performed 55% versus -36.8 for SPY.
I had 69% winning periods with an average return of 4.53% and 31% loosing periods ith a medium loss of
2.3%. Sharpe ratio = 1.37.
I do not think it would be good to include a REIT index. Never iinclude a single ETF you know that it
performed well in the past. This would be a sort of cheating. I could also add Google, Facebook and Apple
to increase the return of my strategy
10 Aug 2013, 12:18 PM
Ian Farbrother
Interesting article, Frank.
I did some work on swapping strategies like this a couple of years ago (using ETFReplay.com). One of the
conclusions I came to at the time was to be very cautious about 10-year backtesting. It was very easy to get great
returns from 2003-2008 if you included Brazil (EWZ) in the mix during that period - and then managed to switch
out before the crash (and, of course, your ILF captured most of that). Going to EDV during the crash certainly
helped!
However, I do find that you have an interesting set of ETF choices here, in particular, using EDV and EPP. I
basically gave up on finding a good strategy in 2011, when 'everything' just seemed to be chopping around,
making it difficult to not get whipsawed every month. It seems as though EDV again came to your 'rescue' then.
I remain cautious about the idea of committing my whole portfolio to this kind of strategy - especially when picking
only one ETF per period - but am thinking of perhaps trying it out on a portion of the portfolio to see how it goes.
In any case, look forward to reading more of your articles. Like rasadagopan I would like to see more info,
especially max drawdown.
Thanks, Ian
09 Aug 2013, 01:58 PM

Author’s reply » You are right. Latin America ILF outperformed the other markets from 2003-2008.
However 2008-2013 none of the markets outperformed in such a way and the annual performance was
47% which is even higher then for the 2003-2008 cycle. I think these markets fit well together for a
switching strategy and the most important is that the Treasury switch works well to avoid market crashes.
10 Aug 2013, 12:18 PM
extremebanker
Nice article Frank:
I track about 20 ETF's. Update monthly to the four to seven that are best performng. Also, I have recently started
going short on the worse performing funds. My criteria are quite simple. Long on funds with high relative strength
and above 200 day average. Short on funds with low relative strength and below their 200 day average.
I alter the long/short ratio as the market goes up/down.
09 Aug 2013, 03:20 PM

Author’s reply » I made a lot of tests going short on the worst ETF of my rotation strategies, however I
could never get good returns. The main problem is that on long terms these ETF's all have a positive
return. If I make a equally balanced portfolio of all 5 ETF's, then I get 14.4% average annual performance
since 2003. This will successfully kill every short strategy.
10 Aug 2013, 12:14 PM
extremebanker
I agree that trying to short as a hedge is too expensive to be beneficial. At least that has been my
experience. I would never short something that is in a long term uptrend and performing reasonably well.
However, there are sectors of the market that are not performing well. Such as Brasil. It can be profitable
to short these sectors when they are in a downtrend. It also can help serve as some hedge against a
falling market. But that should not be the primary reason to short.
19 Aug 2013, 08:41 AM
igggy
Why MDY and not SPY? Also, there is too much correlation between EEM and ILF and even EPP.
The bigger question is what's the real secret behind so much out-performance? (other than data mining) The only
feasible explanation I can imagine is that this strategy will not always beat the market. I would go out on a limb
and predict that it would under-perform going forward due to regression to the mean.
09 Aug 2013, 11:48 PM
Dorky
Everything regresses to the mean, but nothing stays at the mean for long and that's one reason why I
believe the author's work will continue to outperform.
10 Aug 2013, 09:52 AM

Author’s reply » SPY is much less volatile than the other 4 ETF's. MDY is more volatile than the SPY and
fits better. You can not combine well ETF's which have too different volatilities.
Most of the outperformance was generated by escaping all big market corrections and I think that should
also work in the future.
10 Aug 2013, 12:14 PM
Ehoser
I agree, escaping corrections is the key. Buy-and-holders always point out that if you miss the XXX
strongest days your gains will greatly suffer. In The Age Of Deleveraging Gary Shilling points out that you
do much better if you can avoid the weakest XXX days. (But he doesn't say how to do it.) A short version
of his chart for best/worst 50 months on the S&P500 from 1946 to 2010 ( I hope the formatting is
readable):
Best 50 | Worst 50 | Avg Annual Return

long | long | 8.7%
out | long | 3.0%
long | out | 15.5%
10 Aug 2013, 05:18 PM
caedmon
If you only recalculate and choose investments on a monthly basis, how do you escape all big market
corrections? If, say, you recalculate on October 2nd and the market drops 50% from October 4th to
October 12th, how do you avoid that in this strategy?
14 Oct 2013, 12:55 PM
berry
In a drop like that, the market would close and he wouldn't have to.
14 Oct 2013, 01:05 PM
caedmon
The market dropped 40-60% in a matter of days after both 9/11 and the financial crises of 2008. These
things don't happen over months, my question is perfectly valid.
14 Oct 2013, 01:10 PM
caedmon
I should note that I'll be his next customer if this can be avoided/solved to my satisfaction in his system.
14 Oct 2013, 01:11 PM
berry
How did the market drop 40-60% following 911 if it was closed (as it was)?
In total it dropped around 12% and recovered most of that by the end of the month.
Look at the charts he has published on his web site to follow gains/losses.
14 Oct 2013, 01:20 PM
caedmon
Look at this chart and then think your responses over again carefully: http://yhoo.it/1gF5mZZ Why am I
talking to you? The drops are obvious (and I lived them). Who cares if the market closes for a day when
the world is shaken or it reaches the 10% limit or whatever, two days of that and you're down 19%.
Downplaying the risks and how fast the market can move doesn't help anyone.
14 Oct 2013, 02:12 PM
berry
If you can show me a one week fall of 40% to 60% as you stated, you may have a point.
The market fell a total of 14.1% following 911 (where it was closed for a week, not one day), and had
regained about half of that in the same month and almost all one month later.
Actually, it would be better if you didn't talk to anyone until you get your facts correct.
Vist his site where you can see his results.
Annual performance (CAGR) = 42.2% (S&P500=8.3%)

Total performance since 2003 = 4273.6% (S&P500=135.5%)
Volatility = 25.6% (S&P500=20.2%)
Maximal Drawdown = 31.1% (S&P500=55.2%)
Sharpe Ratio = 1.4 (S&P500=0.37)
He admits to a maximum drawdown of 31.1%. If you're looking for total safety, you will never achieve
excellent returns.
cheers
P.S. Your chart won't show for a lot of people.
14 Oct 2013, 02:16 PM
varan
Of course anything can happen, but the maximum drawdown (based on monthly data) of this strategy is
only 19%. There is only one balanced fund VWINX among so many that has a lower drawdown, though
balanced funds are designed for the ostensible purpose of reducing the drawdowns.
14 Oct 2013, 02:28 PM
caedmon
S&P 500:
Week of September 15, 2008: 1255
Week of October 6, 2008: 899
1 Month drop of 39%. 1 Week drops of 11% and 18%. There are numerous other single week and
monthly/inter-month examples in the S&P chart. Yes it wasn't a single day, if you're being a stickler, but
the issue I was questioning still needs to be addressed. I do feel that I've wasted quite a bit of time making
a point that should have been obvious.
That aside, I'm highly suspicious of this strategy due to the way his worst results have been in the last two
years (more easily verified), while the best results were before that when no one had heard of him. If
someone can actually build an algorithm that works similarly and show us the nuts and bolts rather than
selling it to us, that'd be highly interesting!
15 Oct 2013, 02:34 PM
varan
"[ ]his worst results have been in the last two years ..."
That's not quite correct. 2012 was bad, but 2013 has been just fine.
2003 42.66%
2004 21.84%
2005 61.50%
2006 29.56%
2007 24.30%
2008 13.30%
2009 103.33%
2010 10.57%
2011 32.39%
2012 6.51%
2013 21.51%
Using IJJ, IEV, EEM, ILF, EPP and TLT.
If you use MVV instead of IJJ
2007 30.47%
2008 13.55%
2009 109.62%
2010 37.56%
2011 52.06%
2012 17.39%
2013 28.62%
Not bad at all.
15 Oct 2013, 02:50 PM
berry
caedmon said "The market dropped 40-60% in a matter of days after both 9/11 and the financial crises of
2008. These things don't happen over months, my question is perfectly valid. "
I'm still waiting for some verification of this assertion.
caedmon said "S&P 500:

Week of October 6, 2008: 899"
Once again, did you bother to look at his results? His strategy had us in EDV which showed results much
different than what you've reported, You can't take some arbitrary market result and blame it on him, if he
wasn't even in the referenced market. Seems as if his strategy was already working. And he does show a
one month drop of 21.5% (not quite the 40-60% you assert) in Jan 2009. He's not trying to hide anything.
caedmon said " I do feel that I've wasted quite a bit of time making a point that should have been obvious.
"
I feel the same way. Once again, PLEASE check your facts.
15 Oct 2013, 05:20 PM
Elran
Varan did you test it yourself ? or do you rely on Frank's tests.
I did my own tests with 3 month momentum and got much different results , look at my comments in the
bottom of the comments page.
15 Oct 2013, 06:38 PM
varan
I did my own tests. Of course.
15 Oct 2013, 06:49 PM
Learner16
Elran, my tests, using only 3 months total return and the following ETFs (EDV EEM EPP IEV ILF MDY
SHY) give me a 21% CAGR from 2004. There is a 5% loss in 2012 but YTD the results are a 14% gain.
Good, but far from Frank's results (although I did not use volatility at all so there should be a difference).
16 Oct 2013, 09:04 AM

Baboon
What is the permitted drop?
16 Oct 2013, 11:58 AM
varan
Where did you get your data? Testing based on the Yahoo data even with your ETFs yields entirely
different results
2004 18.41%
2005 54.19%
2006 29.80%
2007 24.30%
2008 21.06%
2009 57.19%
2010 13.45%
2011 42.97%
2012 1.54%
2013 3.81%
CAGR 25.95%
Here are some details.
On first trading day of every month: use the adjusted close prices on the last trading day of the previous
month and the last trading day of the month three month prior (Sept to Dec for Jan 1, Oct to Jan for Feb 1,
and so on) to compute the return for each. Buy the one with the highest return at the closing of the first
trading day of the month. Use adjusted close data which takes into account any dividends.
16 Oct 2013, 12:08 PM
Learner16
Varan, I used data from ETFscreen.com. You are right, I made a typing mistake. I meant 27% CAGR
instead of 21% from 2004.
This is the backtest for the last five years (the website would not allow me to publish any other date).
http://bit.ly/19QUfKj
16 Oct 2013, 04:33 PM
Elran
That's the results i get about ~18% CAGR from 2003 , Frank claims " A simple MDY-EDV rotation with 3M
look back is already performing 20% since 2003" this is also not what i got .
Maybe he uses dividend adjusted prices but it will on add 2-3% ,
I cannot see how he gets 41% .
I also tried running my own trend algorithm that work great with sectors it did about 23% .
I did also a crystal ball testing - which means that i back-tested the perfect switch as if i had information
about the future - i got 48%.
That means that if you were a prophet you would get only 48% , so i cannot see how momentum strategy
which has only limited prediction power can achieve 41%.
17 Oct 2013, 05:51 AM
Elran
OK , i simulated on a Monthly basis (not weekly basis like before) , and used end of month data , indeed
the CAGR ~ 29%.
It is strange that simulation 12 weeks momentum is very different from 3 moths momentum.
Maybe the 3 months momentum has less noise in it.
17 Oct 2013, 06:26 AM
mjs_28s
FYI - the market was already trending down leading up to Sept 11, 2011.
The DOW closed at 9,645 on Sept 10, opened on Sept 17 at 9,050 and then reached a mid day low of
7,950 on Spet 21, while closing at 8,280.
So, if you want to count Sept 11, 2011 as the issue then the specific terrorist action from Sept 11 to the
2011 lows was a drop of maybe 15%. Expanding the picture to be more fair and to not chalk up the
market to any single event (recency bias will get you everytime) the previous recent market peak was in
September of 2000 somewhere around 11,400. Previous to that you are going back to the tech crash era
of the DOW near 11,800 in spring of 2000 with the index bouncing all over the place after that.
Sept 11, 2001 was a little impact in an already poorly performing market. If you chart if from the longer
term you will see that by the mid 1999s that we hit somewhat of a peak that would be teased a few times
that lasted for 7 years before being broken for a couple of years and then rebroke (after 2008/2009 crash)
in 2012 with, hopefully no reason to go back that low.
All that aside, using this strategy, or any other that helps you avoid the biggest dips by moving into sectors
or areas with larger potential gives you the net effect of buying low and selling high. If something is
already fairly valued or undervalued it will, much more often than not, be punished less during a pullback
that things that are over valued. Which is why this works over the longhaul
24 Oct 2013, 01:18 PM
varan
No strategy like this can account for singular events like 9/11.
24 Oct 2013, 01:22 PM
mjs_28s
I am not sure if that link you posted was supposed to show the S&P back to the 1950s and starting at
around $16 per share, but it did.
That chart should give you confidence about how insignificant the dips are. Cripes, it ends at 1,800 from a
start of 16. Looks like boring indexing is ok, huh? Throw in even a little avoidance of bad markets and you
will be fine, or just ride out the storm and end with 100X more than what you initially invested or more if
you add to it.
The very evidence that you think supports your argument should actually make you feel ok about the
longer term performance.
If the short-term worries you then use a bucket strategy. Have three buckets. One for short-term 1-3 year
money, one for medium term 5 to 10 year money, and one for long-term 10 to forever money.
You trim gains in the long term bucket to keep the middle-term bucket topped off and you use the middle
term bucket profits that you trim to refill the short-term buckets. This allows you to ride out anything less a
total financial collapse of the world.
24 Oct 2013, 01:22 PM
RM13
With downward market already, wouldn't you have been out of SP500 at the time of 9/11?
01 Nov 2013, 04:28 PM
berry
We're never in SP500, and yes, the strategy was out of the US market (MDY) and in EDV.
01 Nov 2013, 05:28 PM
bud4704
Check charts over 5 plus years. MDY beats SPY.
18 Nov 2013, 08:52 AM
vn888
I have done the 'prophet' experiment as well, and I come to different results. In addition I did a Monte
Carlo simulation.
Here are the results:
Buy best ETF every month: 113.5% CAGR (~4000 times your money after 11 years!)
Buy worst ETF every month: -40.3% CAGR
Buy random ETF every month: 13.5% CAGR (median of 1000 runs)
Min: -1.78%, Percentiles 25/75: 10.32%/17.07%, Max: 28.77%
Symbols: MDY, FEZ, EEM, ILF, EPP, EDV, SHY

Period: 2003-01 - 2013-12
Full monthly listing of best/worst ETF experiment: http://txt.do/1rcl
06 Jan 2014, 08:04 PM
satan2liberals
vn888:"Full monthly listing of best/worst ETF experiment".
S2L: I'm sorry I must have missed the practical implications of your post and link.
Care to spell it out for me?

thx
06 Jan 2014, 08:43 PM
vn888
@satan2liberals:
My post was in response to that of Elran on 17 Oct 2013, 05:51 AM, where 48% CAGR was reported for
crystal ball / prophet testing, i.e. you use knowledge of the future to pick the winner every month.
48% CAGR sounded low to me, so I did the experiment myself and came up with another number
(113.5%). This is useful, since it sets an upper limit for how well you can do (given the monthly period and
the chosen ETFs). Besides knowing how much more potential you have, you could also to some extent
use this to judge whether your system is performing as when you backtested it. E.g. if you normally are at
41%/113% = 36% of perfect CAGR, then you should expect to harvest 36% of the market potential in the
future, the point being that the market potential may vary over time, so you shouldn't expect 41% CAGR
every year.
While I was at it, it was pretty easy to pick the worst ETF to see how bad you could do. Not so useful
maybe...
The final (Monte Carlo) experiment you could call the monkey experiment, i.e. you let a monkey pick the
ETF every month. If your backtest is not significantly better than a monkey, then maybe you should just
toss a coin every month :-)
The link was included for people wanting to verify my results.
07 Jan 2014, 07:43 PM
Doug+Taylor+
Hi vn88 you only took their performance for the previous month to select for the current month in your
113% CAGR ?
Thanks
07 Jan 2014, 08:21 PM
vn888
@Doug12:
No, I don't look back, I look forward (not possible in real life of course): Every end of the month (e.g. 31-
Dec-2013) I look into the *future* to see which ETF will perform the best until the next end of month (e.g.
31-Jan-2014). This is why Elran called it crystal ball testing or being a prophet.
08 Jan 2014, 06:23 PM
sofocal
Varan,
Whre did you get data from for EDV for back test sinse 2004?
EDV was lonched on Dec6,2007 according Yahoo.
Thank you
11 Sep 2014, 11:38 AM
berry
Most backtest for periods before an ETF was introduced simply omit that ETF for periods before it was
introduced, thus for periods before EDV became an ETF, it is not included in the tests.
Some people backtest using what they consider to be an "equivalent", but in most of those cases they
mention what equivalents were employed.
The composition of the funds may change over time also. If you'll look at Franks tests, he formerly used
IEV where he now uses FEZ. It makes for an interesting challenge attempting to duplicate results :)
11 Sep 2014, 12:13 PM
Dorky
Simple approach to a great analysis, and much better than LTCM (it never survive beyond its 3rd year despite
similar 40%+ return per annum) and its bunch of professors.
10 Aug 2013, 09:50 AM
RM13
I think as long as ETFs you use in this allocation were high gainers over period you studied, you win during this
decade long analysis. I can only question overlap between EEM and some of its cousins - Asia ex Japan and
Latin America. So in similar light to others - how about REIT s, commodities, volatility based ETFs (including VQT)
in this allocation? Frontier markets like FRO?
11 Aug 2013, 10:59 AM

Author’s reply » I already did tests including ETF's like
PMNA - PowerShares MENA Frontier Markets
GULF - WisdomTree Middle East Dividend Index
FM - iShares MSCI Frontier Markets ETF
and others.
I agree that these markets probably will be top performers for the next years.
However the problem is that they have very small assets and volume and most only started in 2008 or
later, which does not allow to do backtests. Also some are very volatile.
I will make probably sense to add always one of them together with the top of the 5 bigger ETF's. Also if
you want to get a good return from such a strategy it is not really necessary to have the future top
performers in your strategy. It is much more important that you can escape the next market crash.
12 Aug 2013, 02:50 AM
RM13
Great reply, thanks Frank. I just want to point you towards VQT (SP500 with VIX hedging ETN) and
PHDG (SP500 with VIX hedging ETF) as other possibilities. In its prospectus, VQT is backtested at least
10 years, so that data is available.
12 Aug 2013, 10:54 PM

Author’s reply » I do not like VQT I would never hedge continuously using VIX futures or options. A little
bit of hedging eats half of the SP500 performance. The price is way too high because of the 10%
contango of the VIX front futures.. However I like very much inverse VIX ETFs like ZIV.
Also if you have Treasuries and cash in your rotation strategy you do not need a hedge.
13 Aug 2013, 04:42 PM
RM13
Actually the drag of VIX on performance of VQT/PHDG aren't that great in the right market environment -
if the market is rising and in 'stable' VIX pattern, as little as 2.5% is allocated to VIX holdings. The other
alternative is too buy these hedges only when market starts to correct - dependent on timing and speed of
correction, as little as 5% correction (typically 10%) is needed to activate the VIX safety net.
15 Aug 2013, 11:42 PM

Author’s reply » I tried to replace my (normally) negative correlated Treasuries with VIX ETFs like VXX or
VXZ. However even with very sensitive switching I always lost because of the huge 10% monthly
contango.
19 Aug 2013, 08:12 AM
Mikegyver
Great article and interesting approach. I like it because it is simple and makes sense to me. I am getting ready to
roll over a 401k into an IRA and have been researching the best ways to maximize return with reduced risk. I
would take this over a 60/40 equity to bond strategy.
11 Aug 2013, 11:31 AM
gman1253
Similar to the Scott investment portfolios on the IVY portfolio strategies (Faber). I did a bit of research on the
concept & as Ian stated it's easy to get excellent results backtesting with certain ETFs included in the mix.
Thanks for sharing -
11 Aug 2013, 07:12 PM

Author’s reply » It would be cheating If I would preselect some historically well performing ETF's just to
enhance the result of my strategy. If I compose an equally weighted portfolio of these 5 ETF's then the
back test chart since 2003 is nearly 100% identical to the ACWI ETF (iShares MSCI All-World ACWI
Index). Both would have about 3% annual performance and a maximum draw down of about nearly 60%.
This shows you that here you are trading an all-world selection.
You also see that the strategy did very well in selecting the good ETF's to achieve a annual return of 41%
instead of 3% for an equal mix of the 5 ETF's.
12 Aug 2013, 06:35 AM
Janns2a
Hi Frank,
Great article. Thanks for sharing this.
11 Aug 2013, 10:57 PM
berry
Hello,
Thanks for a fine article. Certainly got me thinking.
You mention you and some friends have used it for a number of years. Does your "real world" experience match
the Annual Performance table? What period does your real world experience cover?
cheers
12 Aug 2013, 08:03 AM

Author’s reply » I am investing in this strategy since 2010, however at the beginning with only a
comparison to cash stop. However I was using country rotation for about 8 years. Today I think that the
return to risk ratio of a country rotation is less good, because sometimes country indexes can crash very
fast. Since 2011 I invest in inverse volatility. I think this is the most rewarding investment you can do at the
moment. I have some strategies explained at my logical-invest.com website
13 Aug 2013, 05:54 AM
shefali1
Thanks for a great Article , I have read so many articles but found yours most promising yet implementable by
individuals.
How would the results be, if we use 1 month performance vs 3 month performance ( to catch trends a bit early to
minimize losses as markets drop a lot faster & quickly) and make a switch to the best performing ETF whenever
we find a leader ( best performer) vs waiting for a fixed period of 1 month?
It will be great to know if you performed this combination .
congratulations for your work.
12 Aug 2013, 10:36 AM
Ian Farbrother
@shefali1. I would also be interested in Frank's thoughts on this.
However, in the playing around I did with similar strategies in 2011, I found that switching based on 1
month trailing performance caused massive 'whiplash' to all the strategies I considered - at least during
the first part of the year, which was highly volatile.
I think it is also important to consider volatility, as Frank obviously does with his Matlab algorithms.
Unfortunately, I wasn't able to really understand how he uses that in relation to the trailing performance
metrics.
12 Aug 2013, 06:15 PM
RM13
Shefali,
I've backtested a number of momentum ETF based strategies on 1 month look back approach - and their
performance is inferior to this. In fact, Frank has probably backtested multiple variations of look back
approaches and I'm betting this is one of the better ones. In fact, 4 or 5 months look back approach - if
right ETFs are chosen - may provide even better returns. I can tell you that strategies with look back over
6 months in my hands have also underperformed. It's about catching the right window - probably
somewhere between 50 and 200 day moving averages.
12 Aug 2013, 10:58 PM
shefali1
HI RM ,
Great to know, is there a back test on the strategy to switch to the leader based on the trailing 1 month
performance Vs 1 month performance at a fixed date.
In essence :
To check the prices for trailing 1 month, if the leader changes based on the trailing 1 month performance,
then switch.
This strategy may be able to capture flow of funds on a more dynamic basis although could be at the
expense of more frequent trades in a choppy market.. but may lead to much better performance in the fast
declining/ rising markets.
13 Aug 2013, 03:41 AM

Author’s reply » Here is a table which shows return in % or different look back periods. M=Month,
W=Weeks
look back return %
2W 5.2%
1M 23%
2M 29.2%
3M 41.2%
3M 1W (peak) 41.6%
4M 39.9%
5M 33.8%
6M 33.5%
8M 27.1%
10M 22.2%
12M 15.6%
The good thing is that the strategy is quite stable. You see this because the above table is smooth and
there are no outlier.
The peak performance is at 13 Weeks. Wen the period is too long then we do not capture the 2008 crash.
When it is too short, every small correction will make you switch which results in a sell low, buy high
behaviour. Anyway you do not want to capture small corrections which are just normal, but you want to
capture long lasting economic cycle trends.
13 Aug 2013, 04:56 AM
Ian Farbrother
Very useful table. Thanks!
18 Aug 2013, 04:52 PM
philkaufman
Agree, this is the type of article us novices want. I too would be interested in the answer to the 1 month
performance question from shefali1, not only to catch more current trends, but also since the seekin alpha
portfolio already calcuates that for us. I don't see how the past 3 month performance would trigger a move to EDV
or SHY unless you had a very long correction and by then you would probably be too late. Maybe I'm missing
something.
12 Aug 2013, 12:33 PM
Ian Farbrother
@philkaufman. At least during late 2007-2008 the markets gave very strong signals of correction well
before the Lehman collapse. I like to use 1-year and 4-year moving averages (proxied by 50-week and
200-week on Google Finance), and the signals were incredibly clear.
Unfortunately, at the time I wasn't paying attention to the markets <sigh> ...
12 Aug 2013, 06:19 PM
rmpalpha
Hi, Frank;
For the purpose of evaluating and ranking volatility, which does this strategy rank higher -- high volatility or low
volatility? ETF Replay considers low volatility to be better, but from the above comments I suspect you may rank
high volatility higher.
Also, I am not clear on "20-day volatility averaged over 3 months." From where can this data be obtained?
12 Aug 2013, 07:31 PM

Author’s reply » It is better not have too much short term (e.g. 20day) volatility. An ETF should not get a higher
ranking only because of a short term spike, so, high volatility is lowering the rank. In fact, the ETF's in a rotation
strategy should have approximately the same volatility. This is not always possible, so using volatility to lower the
ranking can help.
13 Aug 2013, 05:41 AM
Random Number
Thanks for this article! I have pondered, and even tried, various sector rotation strategies, but have never found
one that works consistently for me. The design of your strategy seems likely to prevent some typical problems
with sector rotation.
I have also studied your website, and the other strategies you describe there, and being a congenital sceptic, I
have to ask why you hold silver in your own portfolio. I don't see precious metals listed as a component of any of
your apparently successful strategies.
This is not a "gotcha" question--I'd really like to know.
Thanks.
13 Aug 2013, 07:52 AM

Author’s reply » You are right. However silver is the only investment without some rotation strategy. But
think that I bought it for 19$/ounce which is far below production cost. Also I could buy futures not SLV
and I only had to pay 5% margin. Roll cost is just over 1% per year which is very low. I never saw such a
bargain for a long time! I just have to wait some month.
13 Aug 2013, 04:50 PM
bottom_line_24_7
Hello Frank,
thank you very much for your article - a nice one I looked at several times. However, I do have two questions:
- How do you anticipate a market crash / downturn and step to the sidelines? You state that this essential to a
superior performance.
- Do you factor in any fees and taxes? From your comments I see that there is infrequent trading, but every
rotation can easily cost 100+bps (fees, spreads, ...)
Looking forward to read more from you,

Christoph
p.s.: Are you still in Switzerland?
13 Aug 2013, 08:21 AM

Author’s reply » I do not anticipate a market crash. The marked does it by itself by rising the Treasury
prices and lowering stock prices. And suddenly the strategy will swittch to Treasuries.
If you have a good discount broker the trade fees are not really high if your investment volume is enough.
I think you should have at least 50'000 $ to invest.

I am living in Switzerland near Zurich and have 4 kids and a beautiful wife.
13 Aug 2013, 04:57 PM
Purwakarta
High Frank,
I have seen your critical comments regarding commodities.
Does this also apply for gold?
Best regards
Hans
13 Aug 2013, 08:47 AM

Author’s reply » t is difficult to construct rotation strategies with gold because you have not really
something with a negative correlation to it. The only pair which works a little bit is GLD-UUP (Dollar
Index). However GLD is more like a currency and currencies do not work well for rotation strategies.
13 Aug 2013, 05:06 PM
PC72
Frank,
Looking back at the 2008 crash how did you avoid the huge drop with ILF ? Using the preceding 3 month scenario
ILF would have been the ETF you would have been invested in when it lost close to 80%
Thanks
13 Aug 2013, 04:27 PM

Author’s reply » ILF was only good until end 2007, then MDY was performing quite good for some time,
but starting July 2008 EDV went of like a rocket and showed much better 3 month returns then all other
ETFs in the strategy. So, it was the Treasury ETF EDV which avoided the crash.
13 Aug 2013, 05:12 PM
RM13
One can one up this approach - and granted I don't have the data to prove it - and sell the ETF and go cash when
ETF closes below 200 day moving average. I'm sure Frank could provide us the data.
And finally, I don't think the strategy that involves long term government bonds is a viable one in the environment
of rising interest rates. It was the best hiding spot during 2008 crash, but it is not now, and will not be when the
next crisis comes about. Therefore, my personal view is that 5 ETF approach paired with cash might be better.
13 Aug 2013, 04:41 PM

Author’s reply » There is also a cash stop in the strategy. I just do not consider it as an ETF
13 Aug 2013, 05:34 PM
varan
A good way to spice this up is to use MVV or UMDD instead of MDY. Maximum drawdown (with MVV 22%, UMDD
44%, MDY 17%, using simulated MVV and UMDD for the period 2003-onwards) increases, of course, but the
returns are quite eye popping and probably justified for the mildly speculative non-risk-averse investor. With both
MVV and UMDD you get more than 40% CAGR for 2003-2013 on the basis of ranking that uses just the returns
and not volatility.
13 Aug 2013, 04:55 PM
hermit99
As a perma-skeptic, I wonder if the main motivation of this article is to point people to your website, hoping to gain
customers at $30 or $45 per month?
13 Aug 2013, 05:14 PM
varan
Why do you have to go to his website? You can implement it yourself. It is not that hard. If you can't you
can always go to my website :)
You can even get more than 71% return in back testing, that he advertises on his website, without any
help. (But be ware - any back test that starts from just 2011 comes with a lot of caution.)
13 Aug 2013, 05:25 PM

Author’s reply » To close below 200 day moving average an ETF has already to go down a lot. This is why I do
not like this method to much. A rotation method with a negative correlated ETF like the EDV Treasury is much
better, because EDV goes up very fast if a market becomes troubled and so sometimes the strategy switches to
Treasuries even before you have to realize a loss with a stock market ETF.
13 Aug 2013, 05:17 PM
gman1253
Ahhh the Dick Fabian model my dad used to follow.
I agree with the author, specially in this day & age. Greater volatiliy & the magnitude of the swings makes
the Fabian 200 MA model not as good as year's past. Just my opinion of course.
15 Aug 2013, 03:48 PM
Learner16
Frank, thank you for a very good article. I have a doubt. From what I read, EPP excludes Japan. Does it mean
that the global markets strategy excludes one of the biggest economies in the world? If so, what is the reason?
15 Aug 2013, 03:35 PM

Author’s reply » You are right. Japan is missing. However I do not exactly know why iShares and Co.
always exclude it. Japan seems not to fit in a larger basket of countries with similar economic cycles.
But all together the 5 ETFs are quite a good composition. If weighted 20% each, they replicate nearly
100% the ACWI ETF (iShares MSCI All-World ACWI Index)
19 Aug 2013, 08:20 AM
powerofpi
Frank, the outperformance that you cite is pretty remarkable! It really does sound too good to be true, because it
implies that you have a magic market-timing tool. The natural question is that if market timing were so easy,
wouldn't everybody do it?
Your strategy has two key components: (1) picking a handful of complementary ETFs, and (2) performing a
ranking each month to decide which one to hold. #1 is straightforward, even if it is tedious to optimize the perfect
mix. #2 seems to be where the magic (if any) exists, since this is what performs the function of market timing. In a
comment above, you said you use:
Ranking = 0.7*performance + 0.3*volatility
Like other commenters, I wonder how you came up with these particular weights? I also wonder how you decided
to use a 3-month lookback period. My biggest concern with this strategy is whether or not this is a general recipe
for good performance. Maybe the markets of the past 10 years happened to be the optimal inputs for making your
algorithm look good, but the next 10 years will be mediocre.
Summary: how did you arrive at the details, and what reason do you have to expect that this strategy will
generalize well into an uncertain future?
15 Aug 2013, 05:52 PM

Author’s reply » I found this 0.7/0.3 performance volatility mix by doing an optimisation with the
Quantshare software. Such a mix is necessary if you combine ETFs with different volatility.
Another way to do it without these calculations is by comparing the sharpe ratios of the ETFs. This way
you compare directly volatility weighted return.
27 Sep 2013, 04:30 PM
powerofpi
My concern put differently: given a historical market period, it's not difficult to find a strategy that in hindsight works
extraordinarily well for that period, but perhaps not as well for other periods. How much confidence do you have
that your strategy does not over-fit the peculiar circumstances of the past 10 years? (Over-fitting is a common
problem in machine learning and other backtested quantitative approaches)
15 Aug 2013, 06:19 PM

Author’s reply » I think this strategy is very robust. I have optimized the parameters for a wide range
using Quantshare and you really get stable positive returns for many combinations.
27 Sep 2013, 04:33 PM
extremebanker
Powerofpi:
I would recommend the book "The Ivy Portfolios" as a good read about market timing. Most of those
studies go back over 100 years. There is a section on momentum or relative strength that also shows
serious performance improvement. Most active fund managers are not allowed to use market timing
because it is their job to stay invested. High management fees and having to have some money in cash
can play havoc with performance.
For example, the manager of a large cap growth fund can not buy small cap or value stocks or treasury
bonds during down markets so the manager is really going to have a difficult time outperforming. Plus, he
can not go to 60% cash or anything like that. FINRA would reprimand him for violating the terms of the
fund. When an investor buys an active small cap mutual fund he is buying the stock picking talent of the
manager but no macro changes such as going largely to cash. The investor is going to get small cap
stocks, period.
Market timing does work!
15 Oct 2013, 09:08 AM
IndyDoc1
Frank,
Do you have any thoughts about using the top two performers on a monthly basis instead of one?
18 Aug 2013, 07:31 AM

Author’s reply » It would be cheating if I would just select some historically well performing ETFs just to have a
well performing strategy. However here I could not make a larger market selection.
If I compose an equally weighted portfolio of these 5 ETFs then the back test chart since 2003 is nearly 100%
identical to the ACWI ETF (iShares MSCI All-World ACWI Index). Both would have about 3% annual performance
and a maximum draw down of about nearly 60%. This shows you that here you are trading an all-world selection
and not just a selection of well performing ETFs
You also see that the strategy did very well in selecting the good ETFs to achieve a annual return of 41% instead
of 3% for an equal mix of the 5 ETFs.
I think an All-World index will always have something like a 3% positive performance because of inflation.
So, I am very confident, that such a strategy also works in the future as long as we have this positive trend and as
long as we have markets going up and down in economic cycles.
19 Aug 2013, 08:33 AM
IndyDoc1
Thanks for the reply. I meant to use top two ETF out of the five funds that you choose from. Is this still the same
answer?
19 Aug 2013, 09:20 AM

Author’s reply » The reply was for the question of powerofpi. I don't understand why somtimes a new
questions can move inbetween.
There are some strategies where it makes sense to choose the top 2 or even 3 ETFs. If you have for
example a country rotation strategy with 10-20 countries, it makes sense because it reduces volatility.
However for this strategy with only 5 ETFs the average annual performance goes down from 41% to 31%.
19 Aug 2013, 11:54 AM
gtarolli
Any thoughts on using some volatility index for the overall market to tweak the look back time period. For
example, if the market is quiet, use a smaller window to more quickly catch a movement. And when the market is
noisy, enlarge the window?
23 Aug 2013, 01:03 PM

Author’s reply » The difference is that a volatility index like VIX gives me implied option volatility which is
a forward looking expected volatility. I agree that it could probably be used in a rotation strategy which
only switches between US values like MDY-TLT. I never tested it. The main problem is that I do not have
such implied volatility for other ETFs. However I can calculate historical volatility for all ETFs.
25 Aug 2013, 08:00 AM
RM13
Great article again..
24 Aug 2013, 11:34 PM
powerofpi
Here's one other experiment I ran, and this time there was a bit more history to work with. I think it validates your
GMR approach. I used Vanguard index funds covering the US, Europe, Pacific, Emerging, LT Corporate Bonds,
and LT Government Bonds:
Beginning financial backtests!

Investment pool: [VUSTX, VTSMX, VEIEX, VEURX, VPACX, VWESX]
Max common history: 1996-06-20
Start date: 1997-06-14
End date: 2013-08-23
...
GMR (60,0.8,0.2) annual 60.85%
GMR (60,0.7,0.3) annual 59.9%
GMR (70,0.5,0.5) annual 57.96%
GMR (220,0.7,0.3) annual 55.85%
GMR (60,0.9,0.1) annual 54.48%
GMR (220,0.8,0.2) annual 49.97%
GMR (220,0.9,0.1) annual 49.97%

GMR (60,0.6,0.4) annual 49.32%
GMR (210,0.9,0.1) annual 47.46%
GMR (190,0.7,0.3) annual 43.65%
GMR (180,0.6,0.4) annual 41.68%
GMR (180,0.7,0.3) annual 41.68%
GMR (320,0.1,0.9) annual 41.11%
GMR (190,0.8,0.2) annual 39.22%
GMR (50,0.2,0.8) annual 37.95%
GMR (40,0.6,0.4) annual 37.91%
GMR (50,0.1,0.9) annual 37.67%
GMR (150,0.7,0.3) annual 37.59%
GMR (40,0.9,0.1) annual 37.57%
GMR (30,0.2,0.8) annual 37.52%
GMR (180,0.9,0.1) annual 35.81%
GMR (40,0.8,0.2) annual 35.31%
GMR (180,0.8,0.2) annual 35.28%
GMR (210,0.8,0.2) annual 35.15%
GMR (190,0.9,0.1) annual 35.13%
GMR (30,0.3,0.7) annual 34.89%
GMR (40,0.5,0.5) annual 33.07%
GMR (130,0.4,0.6) annual 32.56%
GMR (60,0.5,0.5) annual 32.41%
GMR (50,0.3,0.7) annual 32.34%
GMR (150,0.8,0.2) annual 30.67%
GMR (150,0.9,0.1) annual 30.67%
GMR (80,0.9,0.1) annual 30.03%
GMR (70,0.6,0.4) annual 29.97%
GMR (40,0.7,0.3) annual 29.25%
GMR (340,0.1,0.9) annual 29.21%
GMR (320,0.2,0.8) annual 29.11%
GMR (30,0.8,0.2) annual 28.83%
GMR (20,0.9,0.1) annual 28.57%
GMR (30,0.9,0.1) annual 28.51%
GMR (330,0.1,0.9) annual 28.18%
GMR (210,0.6,0.4) annual 27.82%
GMR (210,0.7,0.3) annual 27.82%
GMR (340,0.2,0.8) annual 27.55%
GMR (330,0.4,0.6) annual 27.46%
GMR (310,0.6,0.4) annual 27.04%
GMR (70,0.4,0.6) annual 26.88%
GMR (290,0.5,0.5) annual 26.18%
GMR (60,0.3,0.7) annual 26.09%
GMR (360,0.1,0.9) annual 26.01%
GMR (360,0.2,0.8) annual 26.01%
GMR (330,0.2,0.8) annual 25.86%

GMR (240,0.5,0.5) annual 25.69%
GMR (60,0.2,0.8) annual 25.08%
GMR (80,0.8,0.2) annual 25.05%
GMR (30,0.1,0.9) annual 25.01%
GMR (100,0.5,0.5) annual 24.79%
GMR (20,0.8,0.2) annual 24.29%
GMR (90,0.5,0.5) annual 24.09%
GMR (270,0.2,0.8) annual 23.99%
GMR (80,0.6,0.4) annual 23.78%
GMR (70,0.8,0.2) annual 23.69%
GMR (40,0.3,0.7) annual 23.06%
GMR (70,0.7,0.3) annual 22.98%
GMR (40,0.1,0.9) annual 22.76%
GMR (240,0.8,0.2) annual 22.64%
GMR (240,0.9,0.1) annual 22.64%
GMR (220,0.3,0.7) annual 22.62%
GMR (40,0.2,0.8) annual 22.36%
GMR (300,0.1,0.9) annual 22.28%
GMR (300,0.2,0.8) annual 22.28%
GMR (230,0.6,0.4) annual 22.25%
GMR (230,0.7,0.3) annual 22.25%
GMR (30,0.4,0.6) annual 22.24%
GMR (150,0.4,0.6) annual 22.09%
GMR (50,0.4,0.6) annual 22.08%
GMR (150,0.3,0.7) annual 22.01%
GMR (10,0.1,0.9) annual 21.86%
GMR (30,0.7,0.3) annual 21.78%
GMR (260,0.2,0.8) annual 21.71%
GMR (100,0.9,0.1) annual 21.61%
GMR (300,0.3,0.7) annual 21.22%
GMR (50,0.5,0.5) annual 21.02%
GMR (100,0.7,0.3) annual 20.99%
GMR (100,0.8,0.2) annual 20.99%
GMR (50,0.8,0.2) annual 20.94%
GMR (330,0.3,0.7) annual 20.87%
GMR (50,0.7,0.3) annual 20.83%
GMR (270,0.5,0.5) annual 20.78%
GMR (70,0.3,0.7) annual 20.67%
GMR (320,0.3,0.7) annual 20.66%
GMR (80,0.5,0.5) annual 20.37%
GMR (230,0.8,0.2) annual 20.35%
GMR (230,0.9,0.1) annual 20.35%
GMR (30,0.6,0.4) annual 20.24%
GMR (110,0.2,0.8) annual 20.06%
GMR (20,0.7,0.3) annual 19.8%

GMR (250,0.5,0.5) annual 19.62%
GMR (250,0.6,0.4) annual 19.44%
GMR (200,0.7,0.3) annual 19.36%
GMR (40,0.4,0.6) annual 19.31%
GMR (310,0.7,0.3) annual 19.24%
GMR (310,0.8,0.2) annual 19.24%
GMR (50,0.6,0.4) annual 19.05%
GMR (60,0.1,0.9) annual 19.02%
GMR (190,0.3,0.7) annual 19.01%
GMR (190,0.4,0.6) annual 19.01%
GMR (310,0.5,0.5) annual 18.98%
GMR (60,0.4,0.6) annual 18.83%
GMR (110,0.1,0.9) annual 18.19%
GMR (70,0.9,0.1) annual 18.08%
GMR (80,0.7,0.3) annual 18.09%
GMR (10,0.3,0.7) annual 17.94%
GMR (250,0.3,0.7) annual 17.95%
GMR (250,0.4,0.6) annual 17.95%
GMR (270,0.3,0.7) annual 17.84%
GMR (220,0.2,0.8) annual 17.78%
GMR (340,0.3,0.7) annual 17.78%
GMR (220,0.6,0.4) annual 17.75%
GMR (220,0.4,0.6) annual 17.66%
GMR (270,0.4,0.6) annual 17.62%
GMR (320,0.4,0.6) annual 17.59%
GMR (30,0.5,0.5) annual 17.58%
GMR (10,0.2,0.8) annual 17.55%
GMR (50,0.9,0.1) annual 17.52%
GMR (270,0.6,0.4) annual 17.29%
GMR (200,0.8,0.2) annual 17.21%
GMR (140,0.8,0.2) annual 16.79%
GMR (140,0.9,0.1) annual 16.79%
GMR (240,0.6,0.4) annual 16.68%
GMR (70,0.2,0.8) annual 16.62%
GMR (220,0.1,0.9) annual 16.61%
GMR (230,0.5,0.5) annual 16.54%
GMR (300,0.4,0.6) annual 16.27%
GMR (20,0.6,0.4) annual 16.23%
GMR (240,0.7,0.3) annual 16.16%
GMR (110,0.3,0.7) annual 16.14%
GMR (130,0.7,0.3) annual 16.08%
GMR (130,0.8,0.2) annual 16.08%
GMR (250,0.7,0.3) annual 16.03%
GMR (250,0.8,0.2) annual 16.03%
GMR (250,0.9,0.1) annual 16.03%

GMR (350,0.2,0.8) annual 15.4%
GMR (280,0.2,0.8) annual 15.13%
GMR (280,0.3,0.7) annual 15.13%
GMR (270,0.7,0.3) annual 15.07%
GMR (270,0.8,0.2) annual 15.07%
GMR (270,0.9,0.1) annual 15.07%
GMR (20,0.2,0.8) annual 15.04%
GMR (200,0.3,0.7) annual 15.04%
GMR (350,0.1,0.9) annual 15.01%
GMR (360,0.3,0.7) annual 14.85%
GMR (100,0.6,0.4) annual 14.8%
GMR (230,0.3,0.7) annual 14.74%
GMR (170,0.9,0.1) annual 14.68%
GMR (160,0.4,0.6) annual 14.52%
GMR (260,0.3,0.7) annual 14.47%
GMR (260,0.4,0.6) annual 14.47%
GMR (290,0.6,0.4) annual 14.47%
GMR (290,0.7,0.3) annual 14.47%
GMR (290,0.8,0.2) annual 14.47%
GMR (110,0.6,0.4) annual 14.41%
GMR (200,0.9,0.1) annual 14.36%
GMR (360,0.4,0.6) annual 14.32%
GMR (290,0.2,0.8) annual 14.3%
GMR (290,0.3,0.7) annual 14.3%
GMR (170,0.8,0.2) annual 14.24%
GMR (340,0.4,0.6) annual 14.09%
GMR (210,0.3,0.7) annual 14.04%
GMR (170,0.4,0.6) annual 13.75%
GMR (110,0.5,0.5) annual 13.71%
GMR (130,0.1,0.9) annual 13.63%
GMR (340,0.5,0.5) annual 13.55%
GMR (160,0.2,0.8) annual 13.47%
GMR (150,0.6,0.4) annual 13.4%
GMR (160,0.3,0.7) annual 13.38%
GMR (210,0.4,0.6) annual 13.39%
GMR (20,0.3,0.7) annual 13.34%
GMR (20,0.1,0.9) annual 13.03%
GMR (140,0.4,0.6) annual 12.96%
GMR (180,0.5,0.5) annual 12.91%
GMR (280,0.4,0.6) annual 12.85%
GMR (190,0.2,0.8) annual 12.68%
GMR (180,0.3,0.7) annual 12.63%
GMR (340,0.7,0.3) annual 12.57%
GMR (340,0.8,0.2) annual 12.57%
GMR (340,0.9,0.1) annual 12.57%

GMR (110,0.4,0.6) annual 12.52%
GMR (80,0.4,0.6) annual 12.48%
GMR (310,0.4,0.6) annual 12.44%
GMR (20,0.4,0.6) annual 12.43%
GMR (240,0.3,0.7) annual 12.29%
GMR (120,0.8,0.2) annual 12.23%
GMR (100,0.4,0.6) annual 12.14%
GMR (310,0.1,0.9) annual 12.11%
GMR (310,0.2,0.8) annual 12.11%
GMR (310,0.3,0.7) annual 12.11%
GMR (190,0.1,0.9) annual 12.02%
GMR (150,0.1,0.9) annual 11.89%
GMR (190,0.6,0.4) annual 11.85%
GMR (310,0.9,0.1) annual 11.72%
GMR (130,0.2,0.8) annual 11.62%
GMR (180,0.4,0.6) annual 11.59%
GMR (150,0.2,0.8) annual 11.56%
GMR (270,0.1,0.9) annual 11.57%
GMR (130,0.3,0.7) annual 11.52%
GMR (250,0.1,0.9) annual 11.37%
GMR (290,0.1,0.9) annual 11.31%
GMR (260,0.1,0.9) annual 11.28%
GMR (170,0.7,0.3) annual 11.27%
GMR (90,0.6,0.4) annual 11.19%
GMR (130,0.9,0.1) annual 11.16%
GMR (110,0.9,0.1) annual 11.13%
GMR (130,0.6,0.4) annual 11.12%
GMR (350,0.3,0.7) annual 11.12%
GMR (350,0.4,0.6) annual 11.12%
GMR (90,0.7,0.3) annual 11.09%
GMR (130,0.5,0.5) annual 11.09%
GMR (160,0.5,0.5) annual 11.04%
GMR (230,0.4,0.6) annual 10.83%
GMR (290,0.9,0.1) annual 10.72%
GMR (140,0.3,0.7) annual 10.62%
GMR (180,0.2,0.8) annual 10.63%
GMR (70,0.1,0.9) annual 10.58%
GMR (230,0.2,0.8) annual 10.43%
GMR (250,0.2,0.8) annual 10.41%
GMR (150,0.5,0.5) annual 10.32%
GMR (280,0.5,0.5) annual 10.27%
GMR (290,0.4,0.6) annual 10.2%
GMR (200,0.4,0.6) annual 10.16%
GMR (330,0.6,0.4) annual 10.01%
GMR (120,0.9,0.1) annual 9.98%

GMR (360,0.9,0.1) annual 9.99%
GMR (180,0.1,0.9) annual 9.78%
GMR (230,0.1,0.9) annual 9.65%
GMR (340,0.6,0.4) annual 9.47%
GMR (200,0.1,0.9) annual 9.2%
GMR (160,0.1,0.9) annual 9.02%
GMR (220,0.5,0.5) annual 8.97%
GMR (100,0.1,0.9) annual 8.79%
GMR (100,0.2,0.8) annual 8.79%
GMR (240,0.2,0.8) annual 8.63%
GMR (100,0.3,0.7) annual 8.49%
GMR (240,0.1,0.9) annual 8.36%
GMR (360,0.5,0.5) annual 8.29%
GMR (200,0.5,0.5) annual 8.02%
GMR (90,0.9,0.1) annual 7.9%
GMR (330,0.5,0.5) annual 7.91%
GMR (190,0.5,0.5) annual 7.81%
GMR (80,0.1,0.9) annual 7.5%
GMR (120,0.6,0.4) annual 7.46%
GMR (140,0.7,0.3) annual 7.31%
GMR (170,0.6,0.4) annual 7.22%
GMR (200,0.6,0.4) annual 7.11%
GMR (210,0.1,0.9) annual 7.04%
GMR (280,0.1,0.9) annual 7.03%
GMR (90,0.8,0.2) annual 6.9%
GMR (160,0.9,0.1) annual 6.87%
GMR (80,0.2,0.8) annual 6.76%
GMR (210,0.2,0.8) annual 6.66%
GMR (80,0.3,0.7) annual 6.51%
GMR (140,0.1,0.9) annual 6.47%
GMR (140,0.5,0.5) annual 6.44%
GMR (140,0.6,0.4) annual 6.44%
GMR (210,0.5,0.5) annual 6.39%
GMR (160,0.6,0.4) annual 6.35%
GMR (170,0.1,0.9) annual 6.19%
GMR (90,0.4,0.6) annual 6.17%
GMR (10,0.4,0.6) annual 6.1%
GMR (240,0.4,0.6) annual 5.94%
GMR (260,0.6,0.4) annual 5.82%
GMR (260,0.7,0.3) annual 5.82%
GMR (170,0.2,0.8) annual 5.74%
GMR (90,0.1,0.9) annual 5.56%
GMR (200,0.2,0.8) annual 5.41%
GMR (160,0.8,0.2) annual 5.33%
GMR (300,0.5,0.5) annual 5.28%

GMR (300,0.6,0.4) annual 5.28%
GMR (300,0.7,0.3) annual 5.28%
GMR (20,0.5,0.5) annual 5.25%
GMR (120,0.7,0.3) annual 5.25%
GMR (170,0.3,0.7) annual 5.23%
GMR (90,0.2,0.8) annual 5.16%
GMR (110,0.7,0.3) annual 5.13%
GMR (110,0.8,0.2) annual 5.13%
GMR (260,0.8,0.2) annual 5.13%
GMR (260,0.9,0.1) annual 5.13%
GMR (280,0.6,0.4) annual 5.13%
GMR (170,0.5,0.5) annual 5.01%
GMR (90,0.3,0.7) annual 4.98%
GMR (140,0.2,0.8) annual 4.86%
GMR (120,0.1,0.9) annual 4.69%
GMR (120,0.5,0.5) annual 4.47%
GMR (360,0.6,0.4) annual 4.46%
GMR (360,0.7,0.3) annual 4.46%
GMR (360,0.8,0.2) annual 4.46%
GMR (330,0.7,0.3) annual 4.14%
GMR (330,0.8,0.2) annual 4.14%
GMR (330,0.9,0.1) annual 4.14%
GMR (320,0.5,0.5) annual 4.07%
GMR (300,0.8,0.2) annual 4.01%
GMR (10,0.6,0.4) annual 3.47%
GMR (320,0.6,0.4) annual 3.1%
GMR (320,0.7,0.3) annual 3.1%
GMR (320,0.8,0.2) annual 3.1%
GMR (320,0.9,0.1) annual 3.1%
GMR (10,0.7,0.3) annual 3.04%
GMR (120,0.2,0.8) annual 3%
GMR (120,0.3,0.7) annual 3%
GMR (260,0.5,0.5) annual 3%
GMR (350,0.9,0.1) annual 2.88%
GMR (160,0.7,0.3) annual 2.6%
GMR (10,0.8,0.2) annual 2.56%
GMR (120,0.4,0.6) annual 2.33%
GMR (280,0.9,0.1) annual 2.34%
GMR (280,0.7,0.3) annual 2.17%
GMR (300,0.9,0.1) annual 2.01%
GMR (350,0.5,0.5) annual 2.01%
GMR (350,0.6,0.4) annual 2.01%
GMR (350,0.7,0.3) annual 2.01%
GMR (350,0.8,0.2) annual 2.01%
GMR (10,0.5,0.5) annual 1.78%

GMR (10,0.9,0.1) annual 1.49%
GMR (280,0.8,0.2) annual 1.06%
25 Aug 2013, 07:29 AM
GabrielSyme
This was very informative. Thanks much for posting. How do you measure volatility in your backtests?
I've tried to replicate your findings, but thus far I'm seeing returns that are significantly lower than those
you generated above. (I'm using a system that ranks normalized performance and volatility separately
over the lookback period and then weights them. I also find that weights of 0.7/0.3 and 0.8/0.2 seem best.
But the volatility piece seems to detract from returns. Doubtless I'm doing something dumb.)
Thanks again!
30 Sep 2013, 03:14 PM
powerofpi
Frank, I wrote some financial backtesting software to repeat your experiment.
SETUP
-------------------
Dates: 1/22/2009 to 8/23/2013
Investments: [MDY, IEV, EEM, ILF, EPP, EDV, SHY]
Lookback: 10 to 360 days by 10
Performance Weight: 0.1 to 0.9 by 0.1
Volatility Weight: 0.1 to 0.9 by 0.1
Notice that my date range is much more restricted than yours- this is because the earliest date for which your
mentioned investments were trading at the same time is 1/29/2008. I let my experiment begin 359 days after this
(so that all setups had at least a full lookback period at the outset).
RESULTS
-------------------
<Strategy> (<Lookback>,<... Weight>,<Volatility Weight>)
GMR (150,0.1,0.9) annual 52.63%
GMR (320,0.1,0.9) annual 48.24%
GMR (40,0.4,0.6) annual 48.18%
GMR (330,0.1,0.9) annual 47.46%
GMR (330,0.2,0.8) annual 47.46%
GMR (320,0.2,0.8) annual 46.44%
GMR (320,0.3,0.7) annual 46.44%
GMR (350,0.1,0.9) annual 44.48%
GMR (350,0.2,0.8) annual 44.48%
GMR (350,0.3,0.7) annual 44.48%
GMR (230,0.1,0.9) annual 44.22%
GMR (230,0.2,0.8) annual 44.22%
GMR (340,0.1,0.9) annual 44.22%
GMR (340,0.2,0.8) annual 44.22%

GMR (340,0.3,0.7) annual 44.22%
GMR (150,0.2,0.8) annual 43.59%
GMR (40,0.5,0.5) annual 41.63%
GMR (220,0.3,0.7) annual 40.49%
GMR (230,0.3,0.7) annual 39.72%
GMR (250,0.1,0.9) annual 37.44%
GMR (70,0.6,0.4) annual 36.86%
GMR (70,0.7,0.3) annual 36.86%
GMR (80,0.9,0.1) annual 36.17%
GMR (80,0.1,0.9) annual 35.65%
GMR (220,0.1,0.9) annual 35.58%
GMR (220,0.2,0.8) annual 35.58%
GMR (40,0.3,0.7) annual 35.32%
GMR (90,0.9,0.1) annual 34.84%
GMR (60,0.7,0.3) annual 34.82%
GMR (60,0.8,0.2) annual 34.82%
GMR (360,0.3,0.7) annual 34.45%
GMR (60,0.9,0.1) annual 34%
GMR (240,0.1,0.9) annual 33.58%
GMR (240,0.2,0.8) annual 33.58%
GMR (240,0.3,0.7) annual 33.58%
GMR (130,0.1,0.9) annual 33.41%
GMR (40,0.2,0.8) annual 33.3%
GMR (90,0.8,0.2) annual 32.43%
GMR (80,0.7,0.3) annual 31.78%
GMR (80,0.8,0.2) annual 31.78%
GMR (70,0.5,0.5) annual 31.69%
GMR (60,0.6,0.4) annual 31.62%
GMR (70,0.8,0.2) annual 31.5%
GMR (70,0.9,0.1) annual 31.5%
GMR (260,0.1,0.9) annual 30.97%
GMR (260,0.2,0.8) annual 30.97%
GMR (260,0.3,0.7) annual 30.97%
GMR (180,0.1,0.9) annual 30.63%
GMR (180,0.2,0.8) annual 30.63%
GMR (180,0.3,0.7) annual 30.63%
GMR (30,0.2,0.8) annual 30.52%
GMR (210,0.1,0.9) annual 30.22%
GMR (60,0.5,0.5) annual 29.85%
GMR (360,0.1,0.9) annual 29.78%
GMR (60,0.4,0.6) annual 29.1%
GMR (30,0.5,0.5) annual 29.05%
GMR (70,0.3,0.7) annual 29.04%
GMR (70,0.4,0.6) annual 29.04%
GMR (40,0.6,0.4) annual 28.43%

GMR (80,0.5,0.5) annual 28.05%
GMR (80,0.6,0.4) annual 28.05%
GMR (90,0.7,0.3) annual 27.32%
GMR (360,0.2,0.8) annual 27.24%
GMR (30,0.1,0.9) annual 26.48%
GMR (310,0.1,0.9) annual 25.46%
GMR (310,0.2,0.8) annual 25.46%
GMR (310,0.3,0.7) annual 25.46%
GMR (30,0.3,0.7) annual 23.48%
GMR (40,0.1,0.9) annual 23.4%
GMR (120,0.6,0.4) annual 23.23%
GMR (120,0.7,0.3) annual 23.23%
GMR (120,0.8,0.2) annual 23.23%
GMR (120,0.9,0.1) annual 23.23%
GMR (100,0.7,0.3) annual 22.88%
GMR (100,0.8,0.2) annual 22.88%
GMR (100,0.9,0.1) annual 22.88%
GMR (170,0.1,0.9) annual 22.34%
GMR (290,0.1,0.9) annual 21.69%
GMR (210,0.2,0.8) annual 21.48%
GMR (210,0.3,0.7) annual 21.48%
GMR (30,0.4,0.6) annual 21.42%
GMR (80,0.2,0.8) annual 20.55%
GMR (200,0.1,0.9) annual 20.13%
GMR (120,0.3,0.7) annual 20%
GMR (120,0.4,0.6) annual 20%
GMR (120,0.5,0.5) annual 20%
GMR (330,0.3,0.7) annual 18.91%
GMR (30,0.6,0.4) annual 18.88%
GMR (250,0.2,0.8) annual 18.48%
GMR (250,0.3,0.7) annual 18.48%
GMR (200,0.2,0.8) annual 18.3%
GMR (200,0.3,0.7) annual 18.3%
GMR (10,0.4,0.6) annual 18.04%
GMR (190,0.1,0.9) annual 17.54%
GMR (190,0.2,0.8) annual 17.54%
GMR (190,0.3,0.7) annual 17.54%
GMR (50,0.1,0.9) annual 17.41%
GMR (300,0.1,0.9) annual 17.23%
GMR (300,0.2,0.8) annual 17.23%
GMR (300,0.3,0.7) annual 17.23%
GMR (290,0.2,0.8) annual 16.69%
GMR (290,0.3,0.7) annual 16.69%
GMR (90,0.3,0.7) annual 16.21%
GMR (50,0.2,0.8) annual 16.15%

GMR (80,0.3,0.7) annual 16.1%
GMR (90,0.1,0.9) annual 15.62%
GMR (90,0.4,0.6) annual 15.58%
GMR (90,0.2,0.8) annual 15.51%
GMR (90,0.5,0.5) annual 15.49%
GMR (90,0.6,0.4) annual 15.49%
GMR (100,0.6,0.4) annual 15.36%
GMR (80,0.4,0.6) annual 15%
GMR (270,0.3,0.7) annual 14.76%
GMR (170,0.2,0.8) annual 14.65%
GMR (170,0.3,0.7) annual 14.65%
GMR (30,0.7,0.3) annual 13.93%
GMR (20,0.3,0.7) annual 13.55%
GMR (50,0.6,0.4) annual 13.23%
GMR (120,0.1,0.9) annual 12.95%
GMR (160,0.2,0.8) annual 12.93%
GMR (160,0.3,0.7) annual 12.93%
...
DISCUSSION
-------------------
Clearly, my experiment had less historical data to work with, because I constrained myself to a period where all
GMR setups had at least one lookback period of data with which to make decisions. Obviously whatever
experiment you ran must have not considered this.
As a result, I don't think I have enough data to verify your findings. If you look at my results, the top-performing
lookback periods differ by a lot! In fact, they jump around so much that there simply must not be enough data to
differentiate them. That, or the GMR method doesn't really work, but some of the setups simply "got lucky".
I look forward to trying out some setups that have more history. Thanks for getting me thinking!
25 Aug 2013, 07:30 AM

Author’s reply » I am very impressed! Nice back tests. We could work together. How did you program it? One
thing I don't understand is, why you still have a good return with a volatility weight of 0.9 if you include SHY. My
back test would always invest in SHY because the volatility of SHY is nearly 0.
I hope we keep in touch in the future. You can email me to mail@logical-invest.com
25 Aug 2013, 11:56 AM
powerofpi
> One thing I don't understand is, why you still have a good return with a volatility weight of 0.9 if you
include SHY.
I think this oddity was due to the *very* limited price history. If you look at the experiment where I used
Vanguard index funds (so that the experiment had much more data), you'll see that the best weights were
more like 0.8/0.2 or 0.7/0.3, which aligns with your findings.
The software I used is just a Java program which downloads and uses Yahoo Finance data to run tests.
25 Aug 2013, 12:49 PM
TrendXplorer, Contributor
After reading a posting on Tactical Asset Allocation (http://bit.ly/15gTWlU) one of my visitors pointed me at your
article. Your's is a different take and indeed very interesting.
With some effort the thinkscript on my blog can be ported into the DIY strategy you mentioned by "just looking at
the 3-month historical performance and then always invest the next month in the best of the 6 ETFs." including the
ranking you mentioned in the comment section.
Then one would have another "30 seconds a month" system :)
Thank you for sharing, Frank.
27 Aug 2013, 03:34 PM
Kowksi06
It appears that August will be the worst month of the past year. Is it safe to assume that there will be new winner
to invest in for September? Will you be updating the performance of this system as time goes on?
27 Aug 2013, 07:51 PM

Author’s reply » Yes, I plan to have updates from time to time.
29 Aug 2013, 07:00 AM
Kowksi06
Thank You!
29 Aug 2013, 09:24 PM
rabsparks
I just noticed a change from IEV to FEZ, and I wonder prompted you to make the change.
31 Aug 2013, 01:39 PM

Author’s reply » We decided to replace IEV (S&P Europe 350) by FEZ (Euro Stoxx 50). Main reason is
that since quite some time, our bigger asset management customers and also logical-invest is already
buying Euro Stoxx 50 Futures (FESXZ3) instead of the IEV ETF. The performance of the strategy is the
same, but trading Futures is much cheaper. Another reason is that the FEZ ETF trades at more than
double the volume then IEV.
02 Sep 2013, 05:54 AM
jz30
What would be the effect of switching from once a month to once a week? Would that let the noise affect your
performance too much? Would it let you catch market moves faster?
31 Aug 2013, 03:49 PM

Author’s reply » You are right. You would catch many noise down peaks which have nothing to do with
the overlying economical cycle. It would result in a sell low buy high strategy. The performance of the
strategy would go back to from CAGR 41% to 27% for the 10 year back test.
02 Sep 2013, 06:01 AM
jf31
I am paper trading this system based on your monitoring the 5 etfs and choosing the best performer for the past
90 days. As of 08/30/13 MDY went negative. So for Sept. the switch should be to cash as cash has 0 return which
is higher that negative? Is this correct? I like the system. very simple. I want to see how it performs during the
choppy times I see ahead till the end of the year.
01 Sep 2013, 03:27 AM

Author’s reply » I still get +0.5% 3 month average for MDY (close last day of July until close last day of
August). So, we keep going with MDY. But anyway it would be bad to switch now to cash. The worst thing
to do is to sell into a political crisis like the Syria crisis. Remember that we follow economic cycles and do
not try to do a short term market timing. Such a political crisis is normally short lived in contrary to an
economic crisis which can have very long term effects on the markets. By end of September the world will
look different again.
02 Sep 2013, 06:08 AM
Kowksi06
Why are you using the last day of July to the last day of August? I thought we were to use results of the
past 90 days not 30 days to decide what to invest in next.
02 Sep 2013, 05:32 PM
RM13
Frank,
I'm sorry, you don't get to pick cash or MDY - you go with the highest positive return in last 90 days or 3
months - both work similarly almost 100% of the time. If you start selectively cherry-picking investment
options, you no longer have the momentum driven investment strategy one designs.
08 Sep 2013, 04:38 PM

Author’s reply » I look back bout 90 days, but I do this every month at the last trading day of the month.
18 Oct 2013, 09:03 AM
IndyDoc1
I am not speaking for Frank, but if you employ the 5 Global Markets + Treasury then the winner goes to MDY. EDV
performed worse than MDY in the last 3 months. But if you employ the 5 Global Markets + Cash exit then it is
actually a tie between cash and MDY. I would probably choose cash since it outperformed MDY in the last month.
01 Sep 2013, 01:44 PM

Author’s reply » I still get +0.5% 3 month average for MDY (close last day of July until close last day of
August). So, we keep going with MDY. But anyway it would be bad to switch now to cash. The worst thing
to do is to sell into a political crisis like the Syria crisis. Remember that we follow economic cycles and do
not try to do a short term market timing. Such a political crisis is normally short lived in contrary to an
economic crisis which can have very long term effects on the markets. By end of September the world will
look different again.
Here in Europe the Stoxx 50 is just now up 1.7% this monday, first day of september. So I think anyway by
tomorrow cash is out again.
02 Sep 2013, 06:11 AM
varan
My computations suggest that IEV is the ETF for September.
Here are the profits based on adjusted close price data from Yahoo finance for the period from 5/31/2013 to
8/30/2013:
MDY 0.19
IEV 1.28
EEM -6.52
ILF -12.03
EPP -2.34
ZIV 0.99
EDV -9.06
01 Sep 2013, 03:58 PM
varan
Sorry my system does not use volatility. So the above results do not correspond to the author's strategy.
01 Sep 2013, 04:20 PM

Author’s reply » I do not think it makes a big difference if you invest now in US or European market. They are
very close together.
The most important result of the strategy was to prevent you from investing in markets like Latin-America or
Emerging-Markets.
02 Sep 2013, 06:15 AM
Sandra123
Yes but since the end of June emerging markets EEM has started to outperform the US ETF markets by a little bit.
In the last two weeks they have substantially outperformed. Is that what you are seeing?
07 Sep 2013, 11:38 AM

Author’s reply » You are right. Mainly EPP-Pacific and IEV-Europe (or FEZ) begin to outperforme the US
market (MDY). They also have about 40% recovery potential on MDY seen 3 years back. However I wait
until they are no 1 in the strategy ranking. I do not make predictions.
10 Sep 2013, 08:31 AM
Tiziano C.
Dear Frank, thank you very much for sharing your findings.
I have been backtesting the ETF Macro Rotation strategy (5 Macro EtFs, EDV and SPY) on my own using Excel
and VBA. I backtested the strategy using weekly data (I used the closing prices of Friday) with return and
standard deviation as indicator and lookback period and trading period as parameters.
I used as indicator:
Past performance (of last x weeks) – weight*volatility (of last x weeks)
By not taking into account volatility (only past return) the results easily overperformed a buy and hold strategy but
they are very far from your results. My backtest performed best for a trading period of around 5 weeks and a
lookback period in the area 13-17 weeks, which is line with your findings. However what I don’t understand is that
the performance (by the way the period is 2003-2013), for the best lookback-trading period areas, is “only” around
18% annual return and 21% of volatility. I still regard it as a good performance given the financial crisis but I was
just wondering why there is such a difference between my finding and yours.
Here are the best results for 2003 – 2013 (with weight of volatility = 0):
Performance (Lookback period, trading period) = (Return/volatility/sha...
P(13,4) = (19,37/21/0.78/-30)
P(13,5) = (15,94/21.45/0.60/-34)
P(13,6) = (17,56/21.5/0.68/-29)
P(14,4) = (20,5/20/0.87/-23)
P(14,5) = (18,9/19/0.81/-22.5)
P(14,6) = (18,14/21/0.72/-27.4)
P(15,4) = (18,8/20.57/0.77/-37)
P(15,5) = (18,8/21/0.75/-23.3)
P(15,6) = (18,1/21.2/0.71/-27)
P(16,4) = (15,5/21.45/0.58/-46)
P(16,5) = (13,08/21.45/0.47/-37.8)
P(16,6) = (16,5/21.6/0.62/-30)
08 Sep 2013, 09:27 AM
Elran
I got the almost the same results , look at my comment down
02 Oct 2013, 06:37 AM
varan
The most amazing thing about this strategy is that the maximum drawdown is only 20.1%. This is less than most
of the so-called "balanced funds" (only VWINX has the lower drawdown of 18%).
Even the almost equivalent mutual fund version of this strategy using the basket containing FDVLX FIEUX
FEMKX FLATX FPBFX and VUSTX (no mutual fund equivalent to ZIV exists as far as I know) yields a CAGR of
19.7% over the longer period 1991-2013 with a maximum drawdown of 24.5%. Of course the mutual fund strategy
cannot be implemented in practice as most of the funds have a hefty fee for getting out in less than two or three
months. (For comparison with the performance of the ETF basket, the mutual fund strategy returns over 25%
during 2003-2013.)
08 Sep 2013, 05:04 PM
Tiziano C.
In addition to my previous comment where I posted my backtest’s results for weekly data I post also the results for
monthly data. Again, for the period Sept. 2013 – July 2013,by simply looking at the 3-month historical
performance and then monthly always investing in the best of the 7 ETFs I get the following result:
Annual return: 21.4%
Standard deviation: 13.8% (is around 21% using weekly data)
Sharpe:1.3
Drawdown:21.8%
With only the 5 Macro ETFs ad the Cash ETF the results are almost the same.
Again I wonder why there is such a huge difference in results, namely 21.4% vs 34%. Thank you
09 Sep 2013, 09:28 AM

Author’s reply » I think your ranking system is wrong. I do rankings for performance and for volatility and then
add these weighted ranks. This gives me the final rank. Also It is important not to include cash as a stop if
treasuries have a negative correlation. I only do this if treasuries have a higher than -0.1 correlation.
If I test a basic MDY-EDV switch I already get about 23% average annual performance since 2003 and this only
looking at 3 month historical performance.
10 Sep 2013, 09:59 AM
Sandra123
For those of us who are just looking into this system here is a great website where you can plug in the
performance rating and volatility and see what ETF's comes out on top. The 100 largest rankings are ranked for
free - if you want more in depth and smaller etf's you have to pay a fee
http://etfreplay.com
There are a number of articles on the web publishing studies on back-testing and weightings for ETF momentum
strategies. A long in-depth study found that 4 months and 20-days (one month) was the best predictor of future
results. I am still unsure of the best volatility rating however. The Ivy League model seems to use 30% weighting
11 Sep 2013, 10:18 AM
smurf
Sandra,
Thanks for the link. Very helpful.
26 Sep 2013, 01:51 PM
smurf
Frank,
Nice first article......lot to absorb, have bookmarked it for further study.
I've spent a good deal of time in Switzerland (tourist, not expat), using Zurich ($) as a base from which to explore
the rest of the country. Need to spend more time in the south, however, Lugano, etc....
Thanks.
26 Sep 2013, 01:53 PM
varan
Good job. IEV has returned 6.2% to-date this month.
26 Sep 2013, 02:52 PM
Learner16
You can also rank ETFs according to return and relative strenght in
http://www.etfscreen.com
27 Sep 2013, 10:29 AM
jz30
Is anyone using FEZ instead of IEV?
27 Sep 2013, 12:15 PM
rabsparks
I switched to FEZ when I found out. FEZ has tracked along the same lines as IEV.
28 Sep 2013, 09:17 AM

Author’s reply » I use Quantshare to do the ranking and all the optimization. It only costs about 230$ for a lifetime
license including all historical data. You can do simple things like a ranking screener, but you can also program
very complex trade systems.
27 Sep 2013, 04:40 PM
Kowksi06
Which etf does this strategy put us in right now?
28 Sep 2013, 09:25 AM
berry
That's the $30 question. :)
28 Sep 2013, 09:41 AM
varan
Based on the returns alone, ZIV.
28 Sep 2013, 09:42 AM
jz30
I thought ZIV was on his other strategy, with this one here i'm thinking it's FEZ unless my numbers are all
goofed up.
28 Sep 2013, 12:43 PM
varan
You are right. IEV/FEZ is the second one, and so if you do not use ZIV, that will be the way to go.
Without ZIV (and based on rankings taking into account the returns alone) you get 32% CAGR during
2003-2013. That is by itself quite good, but with ZIV you get 44%.
28 Sep 2013, 12:54 PM
rabsparks
I'm running GMR with the seven ETFs having replaced IEV with FEZ. I show FEZ as the number one ETF
for September, and have invested accordingly. Currently the FEZ return is 10.0%.
I am using .7 * (3month return)+ .3 * (volatility) as the equation.
You're right about the difference in the two strategies. ZIV is one of the ETFs on Mr. Grossmann's other
strategy, not GMR.
28 Sep 2013, 04:31 PM
jz30
Isn't the risk much higher using ZIV? Is there a chance that volatility could suddenly spike causing a big
loss?
29 Sep 2013, 12:56 PM
rabsparks
Mr. Grossmann's other strategy is called Global Market Rotation Enhanced Strategy.
29 Sep 2013, 05:50 PM
GabrielSyme
y
How are you measuring volatility? Standard deviation of daily returns during the 3-month lookback? Or
something else? Thanks!
30 Sep 2013, 09:12 PM
Elran
Hi Frank ,
Nice article!
I am using sector rotation strategies as my main investing approach,
and i have also tested many rotation systems over the years.
I tried to simulate your chosen 5 Global markets +EDV with 3 month looking back period but i only got about 18%
annual return.
I tried using also close prices , open prices but the results were the same.
I also simulated the "perfect" strategy that means i bought the ETF
that was about to have the best performance in the coming month as if i had a crystal ball , i did this just to check
what is the potential return of the strategy. I got a result of 48% , that means that if i knew the future the maximum
i could have done is 48%.
Based on my simulations It is unlikely that based on momentum one
can reach 41% , momentum is only one of many forces that influences the price.
01 Oct 2013, 02:30 PM

Author’s reply » There must be some error in your calculations. A simple MDY-EDV rotation with 3M look back is
already performing 20% since 2003. You probably may have one day slippage between sell and buy. There have
been some other readers in this thread which could also backtest about the same performance.
02 Oct 2013, 06:00 AM
Elran
I use end of week close and buy in the next Monday open , I checked my code maybe there is some mistake , but
even a 1 day slippage cannot explain this results.
Did you check if you do not have a forward looking element in your code ? this is quite a common mistake .
All the momentum rotation strategies i know , with or without volatility does not get more the 15% above bench
mark return.
I am also referring to adaptive looking back parameter.
See also Tiziano C comment , we have almost the same results.
02 Oct 2013, 06:36 AM
varan
You have to use prices that reflect dividends. Since you are using open prices, it may be that you are not.
If you are using Yahoo Finance, it is easiest to use adjusted close prices.
02 Oct 2013, 11:01 AM
Tiziano C.
Hello Elran,
I made an error in my computations and the results Mr. Grossmann obtained are correct. Make sure the
series you are using are correct. You can easily test the strategy on etfscreen and get an annual return
above 30%
21 Oct 2013, 10:16 PM

Author’s reply » In my backtests I always sell and buy at the same close price. You have a night inbetween. It is
clear that you can not sell and buy in reality at close, however you can sell and buy very close together the next
day.
02 Oct 2013, 07:31 AM
Elran
you can do a simple test to check your system :
Get 2 random data series and run your system on them , if the system still have big profit it means you have a
problem , since random data should not exhibit any momentum characteristics .
I believe you can use Mathlab to get this random data.
This is a good check for any system to make sure that we do not have any forward looking or any other biases
that distort the back-testing.
02 Oct 2013, 10:16 AM

Author’s reply » You can make a test with SectorSurfer (http://www.sumgrowth.com). You can register for free
and you can insert the 6 ETFs (MDY EEM FEZ ILF EPP EDV). Already with the fully automatic basic settings you
make about 32% per year. Disadvantage of Sectorsurfer is that you do not really know what they are calculating
internally. Also you can not run optimizations on the parameters.
I did all development with Quantshare and I am pretty sure, that there is no error in the backtests. Also I am
playing the strategy since several years and the returns are the returns published on the my website
http://bit.ly/15IIHte, however of course you have to subtract your trading fees and spreads. But this depends of
your broker.
02 Oct 2013, 11:00 AM
Elran
When you say the return are published in your website , you mean backtesting or real time return ?
do you have the real time return from the time you started publishing your signal ?
03 Oct 2013, 10:35 AM

Author’s reply » The return published on the website is the real ETF return. I invested in these strategies since
2010, but my return and also yours will be slightly different. From my published close-to-close returns you have to
substract your trading fees, the ETF spread and the slippage between buying and selling because there are
normally some minutes between the trades.
Also I normally did not buy the ETFs like MDY, but I bought the S&P400 Future which in fact increases my
performance, because it has no management fee. This also can make a small difference. Also I could not always
reinvest exactly 100% of my capital because 1 Future is about 40'000 US$.
But the actual performance is very close to the published performance if you have a good discount broker.
04 Oct 2013, 08:57 AM
Tiziano C.
Hello, I really appreciate this strategy and I have started investing in it with my own funds. I follow exactly the
same process as that one described by selecting the best region out of the ETFs listed by Mr. Grossman. Only
difference is that once I spot the top region (say IEV is the best performer, so Europe is the region) I invest in the
small cap ETF of that region ( MSCI Europe small cap TRN) as the performance is quite higher and it outweighs
the less liquidity. What do you think about that? Same volatility ETFs (the 5 etfs listed by Mr. Grossmann) decide
the region to invest in and once selected the region you invest into the small cap ETF of that region. Second
question: can you reccommend any Swiss bank which has low costs, especially low commissions, where I can
implement this strategy without paying tons in commissions? Thank you
21 Oct 2013, 10:20 PM
jz30
Would the small caps also come with higher volatility
22 Oct 2013, 09:20 AM

Author’s reply » There is absolutely no problem to replace the ETFs with other similar ETFs oft he same
markets. The Strategy more or less just gives you the date when you have to switch from one market to
another. I myself always buy the underlying futures.
However I do not think that this will improve a lot the performance. With small caps you pay better
performance with more volatility. It is just a kind of leverage.
22 Oct 2013, 11:10 AM
cnx669
Frank, it seems like this combination of ETF's may perform very poorly should the Federal Reserve taper QE, as
the stock funds and EDV and SHY would become positively correlated and fall in value. What are your thoughts
on this possibility?
22 Oct 2013, 09:39 AM

Author’s reply » If during tapering the stock markets go up, then tapering is not a problem. If there is a real
financial crisis again, then I am sure that treasuries will automatically become a save haven asset again.
22 Oct 2013, 11:16 AM
rabsparks
I decided to give Mr. Grossmann's strategy a try. I've been attempting numerous backtests since I first read the
article. However it has become apparent to me that I can't replicate his methodology and therefore, although I
have very good results, my results are not the same as the GMR Strategy.
I did invest in FEZ last month before I subscribed to Mr. Grossmanns' strategy, and I am satisfied with its
performance. If GMR replicates anywhere near the historical performance I won't have an issue with my $30 each
month.
22 Oct 2013, 12:04 PM
varan
For those interested (as it seems that many are having difficulty in backtesting) here are the results I get for
various cases by using ranking on the basis of total returns alone. Separate results for 2009-2013 are here as
EDV has been around only since 2009. Since a similar ETF, TLT, has been around since 2003, the results based
on using TLT instead of EDV are also shown.
Case A: IJJ IEV EEM ILF EPP ZIV EDV

2003:2013 CAGR 44.8% Max DD 20.2% Sharpe 1.56 Sortino 3.63
2009-2013 CAGR 53.7% Max DD 20.2% Sharpe 1.51 Sortino 3.66
Case B: IJJ IEV EEM ILF EPP ZIV TLT
2003:2013 CAGR 39.7.8% Max DD 19.2% Sharpe 1.43 Sortino 3.36
Case C: IJJ IEV EEM ILF EPP EDV

2003:2013 CAGR 34.2% Max DD 17.6% Sharpe 1.21 Sortino 2.91
2009-2013 CAGR 29.9% Max DD 17.6% Sharpe .99 Sortino 2.31
Case D: IJJ IEV EEM ILF EPP TLT
2003:2013 CAGR 32.1 Max DD 17.1% Sharpe 1.29 Sortino 2.97
It appears that if you don't use ZIV, replacing EDV by TLT may be better.
The most remarkable result (apart from the mouth watering CAGR) is the low drawn, which is better than the
drawdown of the best of balanced funds which are ostensibly there to provide a low volatility portfolio.
22 Oct 2013, 12:40 PM
Market Map, Contributor

Are these results using the rules from a previous post of yours : " On first trading day of every month: use
the adjusted close prices on the last trading day of the previous month and the last trading day of the
month three month prior (Sept to Dec for Jan 1, Oct to Jan for Feb 1, and so on) to compute the return for
each. Buy the one with the highest return at the closing of the first trading day of the month " ?
22 Oct 2013, 09:12 PM
varan
Yes. Exactly. For back testing with the Yahoo Finance data this is the best and the most convenient way to
do it.
Of course, on Fidelity at least, you cannot give an order to buy at close unless you buy in the multiples of
hundred shares (or use extended hours trading). So there are bound to be some differences between the
realized and theoretical returns.
23 Oct 2013, 10:22 AM

Could you supply a spreadsheet and/or an asset allocation diagram of this ? thanks
23 Oct 2013, 07:02 PM
A few weeks ago I finalized my translation of GMR into "thinkscript". See this (http://bit.ly/17MvaMJ) chart.
Below the equity curve a history line is painted showing the previous allocations (yellow arrow).
Furthermore the chart labels above the equity curve specify the number of months held for each ETF as
well as their contribution to realized profits.
Rotation takes place on the first trading day of the new month. In this example chart capital is allocated to
the top ETF based on values for ROC(63) at EOD on the last trading day of the previous month.
(My backtests show optimum results for 66-68 days).
24 Oct 2013, 04:26 PM
Tiziano C.
thanks for sharing. But what do you mean with "locked profit" in the graph? How can you lock the profit?
30 Oct 2013, 07:31 AM
Admittedly "locked" might not be the most appropriate description. Perhaps "realized profit" is the better
one for reporting the overall profit of liquidated positions.
30 Oct 2013, 08:52 AM
rabsparks
There's an old story about the woodsman who had time on his hands being that winters were harsh and game few
and far between. He decided that he would machine a tiny screw to show the Swiss how good he was. So he
worked on the screw for months and sure enough machined the tiniest of screws.
He then shipped the screw to Switzerland.
He waited many months, but low and behold one day he got his screw back. He carefully inspected the screw and
found that the Swiss craftsmen had hollowed out the screw, threaded it, and then put a cap on the threaded
screw.
Inside he found the tiniest of screws.
22 Oct 2013, 12:55 PM
varan
fables are interesting. real results are much more useful.
23 Oct 2013, 12:47 PM
rabsparks
Ask and ye shall receive: As of today, FEZ (which I bought around a month ago) is showing a 3.8% return.
23 Oct 2013, 03:22 PM
varan
This year FEZ has been better. But since 2003, not much difference between using IEV and FEZ .
23 Oct 2013, 03:44 PM
rmpalpha
It appears to me that many of the above backtests have omitted the author's Volatility component. I have
backtested the GMR strategy on the etfreplay.com website using Momentum (3-months; weight 70%) and
Volatility (20-days, weight 30%), evaluated monthly. The results did not match the author's results exactly because
the above criteria are not the exact criteria of the author, but the results were close enough that I now use the
above strategy as a significant component of my personal investment portfolio.
One question to Frank: On etfreplay's backtests, they Sell at Close the last day of the evaluation period, and Buy
at Open the following day. Is that what you recommend, or do you recommend Sell at Close AND Buy at Close
the last day of the evaluation period (or Sell at Open AND Buy at Open the following day)? I doubt the results
would change significantly, but am curious what you think?
23 Oct 2013, 10:12 AM

Author’s reply » I normally sell the old ETF and buy the new one the first day of the month within very
short time (some minutes). I did backtests and it does not really matter if you do this the first, second or
third day. However sell and buy should be close together with no night inbetween.
24 Oct 2013, 09:44 AM
rmkiefer
In changing from one trading period to the next, I think simple is better. For example, on the first trading day of the
month you can sell the current ETF and buy the new ETF at your convenience and probably get the same results
as Frank's. That's as long as you do so with the buy and sell occuring within a couple of minutes of each other.
23 Oct 2013, 12:19 PM
ETFdude
Frank,
Interesting reading and concept.
While the newsletters work for the hands-on Alpha Seekers,
how about Rent a Strategy offering a family of mutual funds for those of us that are both time and mathematically
challenged?
Merci vilmal
24 Oct 2013, 10:55 AM

Author’s reply » I don't want to open a mutual fund. This is very complicated from the compliance side in
Switzerland. In fact this is more of a hobby for me and I like the contact with my customers.
However if someone in the US wants to open a mutual fund for this strategy, then I am ok to support him.
30 Oct 2013, 12:44 PM
IndyDoc1
And the winner for November 2013 is FEZ with a total return of 13.19% over the last 3 months ( July 31 till
October 31). The second place goes to EPP with 11.3% return over the same period.
02 Nov 2013, 11:18 AM
berry
Calculations seems a bit off. I get 11.4% for FEZ and 10% for EPP.
02 Nov 2013, 11:43 AM
varan
According to Yahoo Finance
FEZ 13.5%
EPP 11.4%
IEV 10%
02 Nov 2013, 01:16 PM
DREW Seeker
I'm just guessing the differences may have to do with:
1. Using the open or the close
2. Using (7/31/13 or 8/1/13) to 10/31/13
I think Varan has also mentioned the concept of adjusted close which would include dividends.
03 Nov 2013, 06:16 AM
rabsparks
I had trouble reconciling my calcs with Mr. Grossmann's numbers, so I emailed him.
In Mr. Grossmann's response I noted, "I think the weightings have changed. Try with 80 20. I use a walk forward
optimization. The weights change over time to adapt to different market environments."
So all of us trying to reproduce his calcs have probably embarked on a fruitless voyage.
03 Nov 2013, 10:11 AM
varan
You should actually thank him for describing the basic idea. Why should he give out all the details?
03 Nov 2013, 10:23 AM
berry
"A fruitless voyage" indeed. I too am a subscriber to several of his reports and can see no possible way
for ZIV to rank where it has given it's 29% volatility, unless volatility somehow becomes a negative force at
some point.
I do believe we'll have to continue to subscribe if we want to indeed "see" his monthly results :)
cheers
03 Nov 2013, 02:49 PM
rabsparks
I have chosen to subscribe to GMR, which does not include ZIV. But with things relating to GMR in
somewhat of a state of flux, I don't see how one can match Mr. Grossmann's formula(s). I know one can
figure out which ETF is Number One, but that's about it.
03 Nov 2013, 03:19 PM
varan
Good for you.
No fruitlessness here. I am doing just fine rolling out my own.
03 Nov 2013, 03:21 PM
berry
Yes, I believe we can mostly replicate the top pick until extreme volatility hits most markets, as it would in
times of turmoil. Once there, I believe we'll likely have to rely on Mr Grossman's "magic formula" to predict
what course we need to follow.
If it can indeed do that, it's well worth the cost.
03 Nov 2013, 04:02 PM
varan
I would be careful with the ZIV. The period of back testing is not very long (just three years).
03 Nov 2013, 04:58 PM
rabsparks
I did thank Mr. Grosssmann by subscribing to his monthly "strategy". Not that I owe you an explanation, but I find
interesting that everyone, or a lot of people, are attempting to replicate his service.
Better read my comment again.
03 Nov 2013, 12:41 PM

Are you trading out of Switzerland using "U.S." ETF's ? What tax do Swiss citizens pay on short term cap gains ?
04 Nov 2013, 09:26 AM
rabsparks
I am not trading in Switzerland, but rather in US ETFs. I have no idea of what the Swiss pay in short term
gains. As far as short term gains are concerned, as long as GMR produces anywhere near the returns
that it has in the past, I'll gladly pay the US government their share of the profit.
04 Nov 2013, 10:47 AM

Author’s reply » We are in the good situation in Switzerland that we do not have to pay taxes on profits from
shares or ETFs. We only pay taxes on dividends. I like to trade US ETFs in CFD form. With CFDs I can keep my
capital in Swiss Francs.
The bad thing is that I have also to pay about 25% US Taxes on dividends of bond ETFs
07 Nov 2013, 10:19 AM
mjs_28s
"We are in the good situation in Switzerland that we do not have to pay taxes on profits from shares or
ETFs. We only pay taxes on dividends"
Seriously? No capital gains tax, only dividends?
I might have to move their with my wife.
07 Nov 2013, 01:10 PM
berry
Not an easy place to get into.
Good luck.
07 Nov 2013, 04:20 PM
Baboon
Aren't CFDs like futures having large cost of carry?
08 Nov 2013, 02:23 PM
mjs_28s
What keeps one out? Educational background? Assets (so you aren't a burden as soon as you get
there)?
08 Nov 2013, 03:53 PM
ithan912
How come you dont mention in this or your BRS strategy that your backtests are longer than the
availability of the ETFs? BOND/CWB not available in 2008 and EDV not in 2003-08...so your backtests
appear to very flawed
20 Nov 2013, 03:06 PM

Author’s reply » Each rotation strategy can be decomposed in some basic sub-rotation strategies. For
the bond rotation you had TLH-JNK (Treasury-Junk bond) since 2007. This sub strategy is the basement
of the strategy and returned 16% annually since 2007. BOND and AGG are just elements to reduce
volatility. CWB goes hand in hand with JNK.
A rotation strategy is never static. You can always add new ETFs if this makes sense.
28 Nov 2013, 03:21 AM
Doug+Taylor+
Thanks Frank, very interesting. I am going to try this with a small portion of my portfolio and see how it works. One
question though, can you post what the 2008 and 2009 picks were to get such impressive market returns? Really
enjoyed reading this.
17 Nov 2013, 06:54 PM

Author’s reply » Just look on my website http://bit.ly/15IIHte. You will find there all investments.
28 Nov 2013, 03:23 AM
Doug+Taylor+
I see EDV was the one that went up in 2008 from your earlier post. Was it held for most of the year?
18 Nov 2013, 12:54 AM
berry
A complete list of all holdings is available on his site - url in his post above.
cheers
18 Nov 2013, 07:35 AM
Doug+Taylor+
Thanks berry
18 Nov 2013, 10:21 PM
Tiziano C.
Dear Frank,
I will subscribe to your strategy. I am thinking to invest all my capital in the GMR strategy and bond strategy, but I
am tempted to invest all in GMR as it delivers a more than double return. I don't want to have other investments
such as funds and in general "buy and hold investments". Do you think a 20% in the bond strategy and 80% in the
GMR strategy is appropriate? Vielen Dank :)
By the way: I checked some monthly performances of your GMR strategy and some seem to high (for example
ILF did not perform 6.25% in December 2006 but something around 3%). But still recalculating all monthly
performances I ended up with 38% annually.
24 Nov 2013, 04:31 AM
rabsparks
I took note of your 12/2006 ILF performance and decided to run ILF through http://etfreplay.com.
According to etfreplay, ILF returned 6.3%. Your analysis may have omitted dividends.
26 Nov 2013, 04:10 PM

Author’s reply » Dont't look only on return. GMRS has double return, but also twice the risk (volatility) of
the BRS
28 Nov 2013, 03:26 AM
berry
Unless "black Friday" brings us something rather dramatic in the market (unlikely for a shortened trading day),
FEZ will again lead the pack for GMR.
26 Nov 2013, 06:25 PM
berry
Apologies - kids came home from school and I got a day ahead of myself. In my world I thought today to be
Thanksgiving.
Anyway, long FEZ.
27 Nov 2013, 08:44 AM
TrickPony
If you considered volatility I think EPP was the pick for November and EMM is the leading horse going into
December.
27 Nov 2013, 12:28 PM
varan
November returns:
EPP -2.9%
FEZ 1.2%
IEV 1.1%
Although one month does not make or break a strategy, looks like you would have been better off in
November if you did not consider the volatility.
27 Nov 2013, 12:32 PM
berry
FEZ was the recommendation for November and will likely be for December unless there's quite a change in his
methodology or the market. EMM isn't a player.
27 Nov 2013, 12:48 PM
TrickPony
Hello Berry
When was EMM removed from the GMR mix?
I see the problem. I was still using the original IEV instead of FEZ.
Wow, what a complete different line up for November & December by just using a different European ETF! I
actually would have made some money in November instead of a 2% loss.
My data from etfreplay.com
28 Nov 2013, 01:10 PM
berry
Sorry, I merely meant EMM wasn't a player in the context of being in the running for a top spot. They are still
included in the GMR strategy.
28 Nov 2013, 09:25 PM
rabsparks
Hey Berry: I think that you have a typographical error. I don't think that EMM is a ETF in GMR. I believe
that you mean EEM.
30 Nov 2013, 02:04 PM
jz30
Yeah FEZ in the last two months and most likely going into December FTW
29 Nov 2013, 09:21 AM
johnmarg
Am I missing something. Where is FEZ listed in the strategy above
29 Nov 2013, 03:58 PM
johnmarg
Never mind I just saw where Fez was added
29 Nov 2013, 09:16 PM
clipit89
Hi Frank
,
What a great article and your follow-ups are greatly appreciated.
I'm still a little confused on the starting date for figuring out which ETF has the best total return for the last three
months. Should I use a starting date of August 30th and the closing prices as of August 30th or use the starting
date of Sept. 2nd and the closing prices of Sept. 2nd. with the ending date of Nov. 29th?
Actually, I'm using ETFreplay.com which gives me the option of using results for the last 3 months. Is that time
period accurate enough? Thank you for your patience with the ones that don't have any experience with this stuff.
Sincerely,
Wayne
07 Dec 2013, 08:57 PM

Author’s reply » I am using closing prces of end month, because then I have time to do the ranking calculations
after the closing. However you can also use opening prices at the beginning of the month. It makes not a big
difference.
08 Dec 2013, 11:28 AM
berry
I have created a Windows based, utility which allows you to rank certain ETF groupings based on performance
and volatility. It provides an approximation of Mr. Grossman's GMR, GMRE and Bond ETF strategies.
It does not provide the same detailed analysis provided in his newsletters. It provides an approximation of
performance using the methodology he has provided in this article.
It is available at http://m8e.com/etf
08 Dec 2013, 12:13 PM
clipit89
Berry,
Thanks for the program. Greatly appreciated. Food for thought: Several years ago I read an article by
Jeffery Sault of Raymond James. He suggested a strategy offered by someone else for our consideration.
The idea was to maintain a 50/50 balance between cash and the investment vehicle on a daily basis. Ex:
10,000 in SPY and 10,000 in cash. SPY goes up 1% giving you 10,100 in it and you rebalance to 10,050
in both. If it goes down same thing. Is it possible to see if such a strategy improve Franks strategy?
thanks again
10 Dec 2013, 10:36 AM
berry
Hello clipit89,
A strategy such as you mention is beyond what I'm attempting to offer, which is merely an easy way to
track the "fundamental" (not detailed) strategy outlined in the articles by Mr Grossman.
I do intend to add additional options for ETF family creation and selection, but for now I will offer an
approximation of Mr Grossman's strategy as he has explained it.
A new beta of the utility is available at http://etfrank.com
My intention is not to clutter Mr Grossman's thread with information pertaining to my utility. You should be
able to contact me via email or via the site. If it appears there is sufficient interest, I will create a forum for
discussing our ideas.
Thanks for your comment.
cheers
14 Dec 2013, 04:20 PM
To accommodate the exploration of similar strategies by interested retail investors, I made an All-In-One
framework to explore studies like, but not limited to GMR, BRS. Since yesterday the suite is available on
my blog.
As an expansion to the work of Frank Grossmann asset allocation can not only be controlled by weighting
momentum and volatility, but by adding portfolio correlation to the equation too.
And finally there is a mechanism for c(r)ash protection in cases where absolute momentum is negative
(actually: lower than cash proxy's momentum).
http://bit.ly/1f6TwK9
05 Jan 2014, 08:17 PM
Marc Cohn, Contributor

Berry, I did this daily rebalancing with SPY and SHY from 2002 to 2014 and the results stink. Total return
of 91% (CAGR 5.5%) bad drawdown, stdev, and sharpe.
06 Feb 2014, 07:52 PM
caribe97
Very interesting writeup.
Has anyone tested this system using the the double or triple leveraged ETF's? For the period above, Sep 1, 2012
- Jul 31, 2013, if I substitute UXJ (x2) for EPP and MIDU (x3) for MDY, returns are 78.6% instead of 28.7%. It
would be interesting to see results over the longer time frames.
10 Dec 2013, 07:21 AM
IndyDoc1
Judging from previous replies, it won't work as leveraged ETFs have much higher volatility. The rotating
ETF strategy require assets that have similar volatilities
10 Dec 2013, 08:47 AM
quoveritas
I am curious about the bar chart you show with a and b indicated in red and blue. I feel I must be
misunderstanding something as it appears that during the b (investment) period the etf represented is pretty flat
which would say no gain??? Also I wonder if there are higher volume alternative to some of your lower volume
choices that work almost as well, especially an alternative to ziv in your linked article?
28 Dec 2013, 06:16 PM
Seattle Pilot
Frank, before subscribing a strategy, I have a question. Do you think all the subscribers will be able to exit a low
volume ETF like ZIV, even more people will exit by looking at the VIX term structure. Besides, if you were to place
orders, would you place a market order - say for $100K. I wonder where that will end up.
04 Jan 2014, 01:47 AM
satan2liberals
seattle pilot:"Besides, if you were to place orders, would you place a market order - say for $100K. I
wonder where that will end up. "
S2L: Frank already answered the specific question but

"market orders"!
Quite frankly for myself I cannot imagine a scenario that is appropriate to use a market order in , EVER.
It screams "let me out/in NOW! , Please rip me off while you're at it".
imho: market orders are only good for panicked situations or part of a simultaneously executed option
strategy ( even then I'm not sure it's a wise choice).
06 Jan 2014, 01:34 PM

Author’s reply » I can not guarantee this, but Velocityshares is a good market maker. I used to trade up to
500'000$ in ZIV and never had a problem to sell. However I never had to sell during a really turbulent market.
You can also short VXZ. It is basically the same as going long ZIV.
Best is to short the VIX Future month 5 or 6. CBOE opens 2.5 hours before the stock market and these 2.5h are
normally quite calm.
I would not place market orders. Better place orders with a limit at bid price.
06 Jan 2014, 11:00 AM
designshoe
I tend to agree with the author except those results 30-40% are far too high to be realistic, even the best trend
follower hedge funds cannot achieve over 13% over 10 years.
the main reason is the black swans, like 2002, 2008, 2011, and who knows when.
looking at ACCU for example, the only did 10% total return in 2 years
16 Jan 2014, 10:41 AM
rabsparks
You commented that "those results 30-40% are far too high to be realistic", but you failed to indicate your
back test results. I went through the author's annual results with my own back test, and either got his
stated returns or ones that were at least comparable.
Please explain your conclusion as to why you compared the author's results to "the best trend follower
hedge funds".
And finally, I show exemplary returns for both 2008 (55.40% author vs. 40.27%) as well as 2011 (89.40%
author vs. 58.05%), having no data for 2002.
I do believe that the author's ETF picks are absolutely critical to the approach.
16 Jan 2014, 03:43 PM
designshoe
yes, but doing this ETF rotation in real time gives you a fund like
ACCU or
ICCIX or
ICSIX
returns far worse than S/P 500
better to use the contrarian value approach.
I am currently in EWI, GREK for the second year now
19 Jan 2014, 03:01 AM
satan2liberals
DS:"yes, but doing this ETF rotation in real time gives you a fund like
ACCU or
ICCIX or
ICSIX
returns far worse than S/P 500 ''
S2L: How do you figure that, I don't see it?
19 Jan 2014, 04:40 AM
jz30
Seems like shorting VZY or even UVXY is a sure path to some profits on the short side
19 Jan 2014, 08:36 AM
TrickPony
Anyone want to venture a guess for February?
31 Jan 2014, 03:27 PM
IndyDoc1
it looks like MDY is the winner
31 Jan 2014, 04:28 PM
varan
Not if you include FEZ.
31 Jan 2014, 04:39 PM
IndyDoc1
On absolute term basis, MDY outperformed FEZ from 11/01/2013 till 1/31/2014. So, MDY wins
31 Jan 2014, 06:19 PM
berry
I have mdy at the top. And after even more back-testing, my utility will be even better in a few days.
Personally, the market is getting a little too unpredictable for my liking.
http://etfrank.com
31 Jan 2014, 06:23 PM
varan
My bad. I was looking at the wrong month.
MDY it is by a substantial amount.
Of course, if you include ZIV, it is that.
31 Jan 2014, 07:03 PM
berry
Yes, ZIV appears to be the winner in GMRE, however it's poor current month performance, along with
very high volatility + what I consider to be a somewhat "unpredictable" market, has me on the sidelines as
far as ZIV is concerned. For the "followers", I hope I'm wrong.
http://etfrank.com
31 Jan 2014, 07:43 PM
berry
Upon looking at the results a bit closer, depending on Mr Grossman's volatility threshold, ZIV may not be
the GMRE winner for January. Based on my latest calculations, which includes a negative factor for very
high volatility, MDY will win for both GMR and GMRE. Guess only us subscribers will know for sure.
31 Jan 2014, 08:54 PM
Kowksi06
And the winner is...... EDV
01 Feb 2014, 12:11 PM
TrickPony
EDV & MDY are polar opposite.
I Think Berry's app shows MDY while ETFreplay has EDV.
If either of these etf's has a substantial increase the other will get killed.
I hate when mechanical investing gets difficult.
01 Feb 2014, 03:02 PM
berry
Don't trust the current version of berry's app. Wait for a revision.
01 Feb 2014, 04:36 PM
TrickPony
Thanks for the response
Kurt
01 Feb 2014, 04:56 PM
berry
There is a "better" version of the utility available.
There appears to be some confusion regarding the utility. The utility is NOT a substitute for Mr
Grossman's newsletter service. We merely provide a utility which may help you test or validate various
strategies for yourself. For example, if you want to use a strategy based strictly on performance, as Mr
Grossman mentioned above, this utility should allow you to do that. In actuality, the utility should show you
that differing values for performance and volatility do make a difference in overall performance. However,
if you want to match the recommendations of Mr Grossman's newsletters, you should subscribe to one or
more of them. We offer no guarantee that this utility, as configured, will match the newsletter performance.
In all probability, it will not. The newsletters offer a much more thorough analysis of the market and the
rationale for the associated recommendations. This utility does not make recommendations.
In any event, we are all indebted to Mr Grossman for his insite and analysis, and particularly, getting us
started with an excellent foundation of ETFs.
04 Feb 2014, 09:19 AM

This rotation system is truly amazing. There are very few rules-based systems that perform this well, and it will be
interesting to see how it performs in the future.
I have generated a variation of Mr. Grossman's GMR that does not use ZIV, uses only monthly data, and provides
42% CAGR since 2004, with a max drawdown of only 16.1%. The annual stdev is 22.56% and the Sharpe
(relative to SPY) is 1.53. Relative to cash, it is 1.86. This version has beat the S&P every year since 2004, even in
the bull markets.
Some other features and differences:
(1) I use SSO instead of MDY because the volatility is much closer to the other ETFs than MDY. SSO is more
liquid than ZIV and, believe it or not, still not as volatile despite being leveraged.
(2) I also use a cash-stop (SHY) based on SSO-EDV correlation over a 4-month period. Using 3 months or 5
months also worked well, showing that the stop is robust, but 4 months gave the best performance. It sent us to
cash only 6 times since 2004 but it worked great. For example, a cash signal was provided Dec 31, 2013...
(3) I rely on 6-month volatility, not 20-day. This makes calculating the signals easy using the data from Yahoo and
an excel spreadsheet.
(4) I use a standard distribution to normalize performance and volatility, rather than an arithmetic ranking from 0-1.
(5) I rely on the log of 3-month performance, rather than the 3-month performance itself. This affects somewhat
the ranking process and makes sense when you consider how volatility is calculated.
Returns are below (vs SPY dividend-adjusted):
2004 16% (11%)

2005 80% (5%)
2006 19% (16%)
2007 22% (5%)
2008 57% (-37%)
2009 51% (26%)
2010 50% (15%)
2011 77% (2%)
2012 23% (16%)
2013 44% (32%)
Sadly, I received a signal to buy SSO on January 31, 2014. Monday was not kind. But today is looking nice. Who
knows what this young month of February will bring?
The system will go to EDV at the end of the month if everything stays flat.
04 Feb 2014, 04:59 PM
satan2liberals
To MCohn:
Very interesting.
I'm not asking for your secret sauce but would you be willing to explain
the math/process required to do steps 4 & 5 in excel here or via PM?
Thanks
04 Feb 2014, 05:28 PM

Satan, I PM'd you some secret sauce. There's more at my instablog.
http://bit.ly/1b2F0BI
For step 4, you take the natural log of the monthly price returns ln($month n / $month n-1). Then you just
add up the logs of the last three months.
For the standardization, I just use the Excel standardize function. Bill Gates has already done all the hard
work for us.
05 Feb 2014, 02:33 PM
Tiziano C.
Thank you for your comment. Can you explain though why you say that SSO has volatility that is close to
the other ETFs compared to MDY? I computed on Excel the annualized volatility both for daily and
monthly returns and results say that SSO's volatility is less closer (than MDY) to the medium volatility of
the other ETFs. Thanks
09 Feb 2014, 12:31 PM

Not sure what to say... maybe we are calculating volatility differently?
For me, I make a column of the natural log of the monthly returns. Then I take the stdev of that column
and multiply by sqrt(12). That's it.
For SSO, this volatility is 29%. The others are in the mid-20%. MDY is in the teens.
10 Feb 2014, 10:50 PM

More on this, it might not simply be that absolute volatility is what needs to be common among the ETFs.
In some sense they also need to trade on roughly the same cycles, in this case a fiscal quarter (3
months). But this likely does not account for the difference between SSO and MDY, as these likely trade
with the same patterns.
11 Feb 2014, 08:47 AM
berry
Since SSO is a leveraged ETF, it would not be unexpected that it's volatility be higher than a similar base index
ETF. No surprise there.
11 Feb 2014, 09:17 AM
The performance of MCohn's version is truly amazing. Very much appreciated, MC.
http://bit.ly/MpPxvn
(with settings for R/V/C: 3m/6m/4m)
Is anyone able to provide an explanation beyond the "turn of the month" effect?
16 Feb 2014, 09:56 PM
varan
You get almost close to what you have (~40% CAGR) with the simple system (without volatility) for MVV,
FEZ, EEM ILF EPP and EDV (with MVV replaced by its synthetic counterpart for 2003-2006 and EDV
replaced by TLT for 2003-2006).
17 Feb 2014, 03:37 AM
IndyDoc1
Varan,
What is the look back period? Is it still 3 months?
18 Feb 2014, 01:43 PM
Thanks varan.
In the spirit of Breaking Bad: how do you cook your synthetic prices?
Are the result with the correlation cash stop at 0.75?
17 Feb 2014, 05:04 PM
varan
1. To do this you only need synthetic daily returns and not prices.
2. a is the adjusted close for any day I for MDY, and b is the adjusted close price for the previous day, the
return for MVV for that day I would be 1+2(a/b).
That is it.
17 Feb 2014, 05:19 PM
varan
Sorry I goofed in writing the above. I use 1+2(a/b-1) for the synthetic returns for MVV if a is the today's
adjusted closing price for MDY, and b is yesterday's adjusted closing price for MDY.
17 Feb 2014, 06:14 PM
No problem, I got the jist of your first reply. Thanks for explaining.
18 Feb 2014, 04:03 PM
pipbel
Guys, why not using VGK instead of FEZ and VWO instead of EEM ?
Volatility and all else equal this gives you additional divident plus 3.7% !
Can someone make the calculations in this scenario ?

Thanks
20 Feb 2014, 07:07 PM
pipbel
I don't find it appropriate to rely only on this GMR strategy for the entire portfolio. Too much emphasis on only 1
factor - momentum.
Can be lets say 1/4 to 1/3 of total PF.
If combined with 1/4 contrarian approach will be better I believe.
Can be added also sector rotation section.
And why not juicy divident section as well to enhance total PF divident yield.
21 Feb 2014, 01:26 AM

jz30
You could have 1/3 all of Mr Grossmans strategies (including bond etc), 1/3 Dividend growth, and 1/3
volatility products (short volatility, selective longs)
22 Feb 2014, 11:35 AM
jz30
We'll maybe it's good to leave a couple 2-3 percent for speculation (appl, tsla, amazon etc)
22 Feb 2014, 11:36 AM
TrickPony
Anyone using real money in this system? I jumped in buying EDV based on total three month returns. In hindsight
I wished I would have stayed with FEZ.
Now that's it about 22 days into the month would Mr. Grossmann let us now if his system also picked EDV?
21 Feb 2014, 12:37 PM
jz30
I've been using it since october and fez was doing good until the end of january, pretty much ate up all the
gains.
I run the numbers manually and it was a close call between EDV and MDY but had MDY slightly higher,
so I went with that and it's been going good so far.
I think you may have to use some judgment in these situations, since we saw the massive selloff at the
end of jan it would be logical that all equities would be temporarily depressed, which could have increased
your weighing to MDY from EDV.
I just saw the emerging market debacle as an excuse for the market to sell off since it was extremely
overbought, just wish i had gotten out of FEz a little earlier
22 Feb 2014, 11:39 AM

Mine picked EDV today. I sold SSO from February (8.5% gain) and have my fingers crossed for EDV in
March.
I and several of my friends are using real money.
28 Feb 2014, 05:31 PM
jz30
Yep I was just figuring the numbers and was surprised to see that. We'll better just follow the system see
how EDV does for the month, maybe we got some sort of mini correction like in February and that'll surely
give it a boost.
01 Mar 2014, 12:30 PM
berry
Yes, I've been following the strategy and using real money. And yes, Y-T-D returns have been
disappointing, but we have to expect that from time to time. On average I hope for and expect reasonable
returns using this strategy, but it is possible for the strategy, or any, to experience periods of negative
return.
For those who have been asking, there is a "revived" version of the ranking utility on the website, if you
want to experiment with a momentum-based strategy. But don't expect it to achieve Mr Grossman's
returns. It's more to pique your interest in the overall process.
http://etfrank.com
09 Mar 2014, 02:40 PM
extremebanker
I am running a similar strategy on my instablog and with real money.
http://bit.ly/1hdnJIm
27 Mar 2014, 09:24 AM
pipbel
EDV is too illiquid. This is from today :
Volume 9,089
You can't really rely on trading with such volumes.
Should be used TLT instead of it. Am I right ?
21 Feb 2014, 05:47 PM
varan
Without volataility IJJ, FEZ, EEM, ILF, EPP, TLT
2008 13.28%
2009 103.33%
2010 10.63%
2011 38.16%
2012 8.62%
2013 22.25%
2014 1.15%
CAGR 28.79%
On replacing TLT by EDV
2008 50.65%
2009 92.44%
2010 13.44%
2011 48.44%
2012 7.93%
2013 5.49%
2014 -9.64%
CAGR 30.05%
21 Feb 2014, 05:57 PM
TrickPony
It seems like every system works great until you put money into it. Must be another Murphy law thing.
22 Feb 2014, 02:42 PM
jz30
yeah i'm definitely hoping that's not the case. Need more than just a couple of months to come to a
conclusion. Seems to me a big part of investing involves patience, and waiting around, wish the results
could be acquired as quick as the backtest!
24 Feb 2014, 10:33 AM
designshoe
good luck with all the B/S trading back and forth
you'll just waste time and a decade will go by and you'll still underperform
nothing works well in real time except LT value investing.
you get to define what value is.
01 Apr 2014, 08:50 PM
pipbel
No one is using GLD in his research. Any idea why ?
It has great momentum now.
Thinking to incorporate it somehow in some of the strategies...
23 Feb 2014, 03:44 AM
rabsparks
From August 13, 2013
Author’s reply » t is difficult to construct rotation strategies with gold because you have not really
something with a negative correlation to it. The only pair which works a little bit is GLD-UUP (Dollar
Index). However GLD is more like a currency and currencies do not work well for rotation strategies.
27 Feb 2014, 12:12 PM
apple2017
Hi Frank,
So, your strategy holds a position for at least 1 month. What if a sharp market meltdown happensin the very
beginning of it? Will you be willing to wait till the holding period's end? If not, what your exit criteria will be?
This kind of disaster planning is very important for any trading method, IMHO.
28 Feb 2014, 08:02 PM

Author’s reply » I just follow the monthly cycles because I backtested every type of stop-loss level and always the
resulting performance was less good than without. However the strategy just gives you a ranking. It is not
forbidden to think and act yourself.
Also I don't invest only in this strategy. I am quite diversified with bonds (junk, convertible and treasury) and also
precious metals (silver). This will normally offset losses in this strategy.
11 Mar 2014, 06:20 AM
ikkyu
Brilliantly simple and effective. Momentum is a legitimate market inefficiency.
Thanks for your article.
Cheers from Osaka,
john
27 Mar 2014, 09:02 AM
Market Trends Investor, Contributor

Frank, what is your switching mechanism to EDV and SHY? Is it a moving average? Is it end of month?
21 May 2014, 12:13 PM
AlexanderHorn
Noted that Frank is using Quantshare for his backtesting. People interested in jointly developing a strategy based
on Franks insights, please feel free to joint in: http://bit.ly/1t8ULtw
25 May 2014, 05:56 AM
Baboon
How much does it cost to use? I could not find it anywhere on the website.
25 May 2014, 07:43 AM
AlexanderHorn
Here is pricing info ($245), first month free with full functionality: http://bit.ly/1guFX8b
Glad if you join in!
.. Not trying to make any commercials here, just a happy user ...
26 May 2014, 03:08 AM
designshoe
just another subscription seller for a useless market timing strategy
26 May 2014, 09:04 PM
satan2liberals
shoe
just another subscription seller for a useless market timing strategy
=============
I don't consider that very fair or accurate.
The author has gone to considerable length to educate readers and suggest non-subscription alternatives.
Also I found the concepts contained rather profound.
26 May 2014, 09:47 PM
rabsparks
I disagree.
1. Mr. Grossmann put enough information out there (seekingalpha) to allow anyone to get close to his
GMR returns. I am running GMR by using the Yahoo published information.
2. His market timing strategy, GMR, (I backtested it for over ten years) has produced the stated returns +/-
a reasonable "error". Likewise other readers have run backtests of the strategy and gotten the same "in
the ball park" returns.
3. Most important, I have an excellent return over the past six months with GMR.
27 May 2014, 11:31 AM
satan2liberals
rab
3. Most important, I have an excellent return over the past six months with GMR.
========
OOOOhh color Me jealous, what kind of returns are we talking about here?
27 May 2014, 11:46 AM

Anybody want to share the switching mechanism to EDV and SHY? Is it a moving average? Is it end of
month?
27 May 2014, 11:51 AM
varan
This year, over 10% YTD without ZIV. Not filthy good, but much better than the market.
27 May 2014, 12:04 PM
AlexanderHorn
@rabsparks,
thanks for your comment!
Indeed not trying to sell anything here, but due to the lot of comments above thought it might be a good
idea to invite people to backtest jointly..
Franks postings have been very helpful indeed, and results of this strategy are great... so why not work
together on this?
btw, also sent a private note to Frank to appreciate and invite him..
27 May 2014, 03:44 PM
TrickPony
Up 10% for this year, what time frame are you looking at?
It is down about 2% from January.
28 May 2014, 07:56 AM
ikkyu
I show Global Market Rotation is up 2% in 2014 as of today (5/30). Maximum Yield Rotation is up 6.3%.
29 May 2014, 07:42 PM
varan
The difference may be due to the fact that I do not consider volatility at all for rankings.
Basket: IJJ IEV EEM ILF EPP TLT
IJJ 1/01/2014 TO 2/01/2014 -4.05%

IJJ 2/01/2014 TO 3/01/2014 7.31%
IJJ 3/01/2014 TO 4/01/2014 2.67%
TLT 4/01/2014 TO 5/01/2014 4.08%
ILF 5/01/2014 TO 5/29/2014 1.08%
2014 -4.05% 7.31% 2.67% 4.08% 1.08%
NET 11.21%
29 May 2014, 07:54 PM
berry
However varan, your portfolio is different - GMR is EDV, MDY, FEZ, EPP, EEM, ILF and it has different
monthly selections for 2014 than you do.
GMR got off to a bad start with FEZ and has only recently recovered.
So, not exactly an apples to apples comparison. ;)
30 May 2014, 12:03 AM
rmpalpha
Varan: Since your portfolio components and filtering criteria are different from those of the GMR strategy, I am a
little confused regarding why they are posted on this thread. On the other hand, since the results appear to have
been better than those of GMR so far this year, I am interested in exactly what your strategy is. You list the six
equity components, and stated that you do not take volatility into consideration, but you did not specify the
complete strategy. Please state the complete strategy (i.e. - equities, ratios, lookback periods, etc.) so that we
may backtest the strategy more comprehensively. Also, is the strategy static or dynamic (i.e. - does the strategy
use the same equities, ratios, lookback periods, etc. continuously, or does it change these items over time)? If
dynamic, what are the criteria for changes?
30 May 2014, 09:53 AM
varan
They are not that different, if you look at what they are in greater detail.
IJJ Mid Cap Value

MDY Mid Cap
IEV iShares Europe

FEZ SPDR Europe
TLT iShares 20 Year Treasury Bond

EDV Vanguard Extended Duration Treasury Index
IJJ, IEV and TLT are available to me without transaction cost.
Everything else is the same, except that I do not use volatility for ranking. Only the return for the prior
three months and update on the first trading day of the month. Simple.
30 May 2014, 10:03 AM
berry
Yes, I realize they are not that different conceptually, however in practice, they have lead to different
selections for 2014, thus the difference between your very good performance and GMR's not quite as
good. re: TrickPony's earlier comment regarding the GMR strategy performance YTD.
cheers
30 May 2014, 10:48 AM
varan
great.
I think that for the purposes of comparison I had previously detailed the results of various back tests,
including one for this set of ETFs, in one of the comments.
It may well be that the difference arises due to the exclusion of volatility in my test. Although a composite
of volatility and return has been advocated at various places, I do not quite buy that. The simplest reasons
for my discomfort are (a) it complicates the strategy and (b) to get just the right combination of volatility
and return you may have to optimize, which in my book should be avoided unless absolutely essential as
it may lead to over-fitting with the consequent large generalization errors.
30 May 2014, 11:11 AM
berry
varan,
I believe the differences between your strategy and GMR are more likely due to "timing" than volatility.
While both strategies are very similar as far as strategy is concerned, the timing of the actual numbers for
performance can lead to the selection of a different fund for a month. Thus GMR's selection for FEZ for
January instead of MDY, which would have aligned better with your IJJ and led to a smaller loss to begin
the year.
GMR has also been in EDV for Feb, Mar and partially (a new wrinkle) Apr.
I've run the GMR strategy with and without volatility and either way, it leads to some of the selections
being different than the equivalents in your portfolio. I believe the way timing is handled is a bigger factor
in the differences than volatility in most cases.
It's almost impossible to make monthly comparisons from different market baskets even though they may
be "equivalents". That's not to say one is better/worse than the other, just that there will be differences at
the micro level.
cheers
30 May 2014, 12:27 PM
AlexanderHorn
Including volatility and standarization, following M. Cohn ETF selection:
Month start Ticker log return Cum

Jan-14 SSO 8.68% 108.68%
Feb-14 EDV 1.91% 110.59%
Mar-14 EDV 2.90% 113.49%
Apr-14 ILF 1.15% 114.64%
May-14 ILF -1.76% 112.88% (till 5/30)
Net 12.88% (largely due to drop in SSO in Dec 13)
The inclusion of volatility seems reasonable to me as it normalizes across ETF as stated by Frank and Marc. To
add confusion, both neg (Frank) and pos (Marc) normalization works quite fine. Agree with varan on potential
over-fitting and complication.
30 May 2014, 11:44 AM
varan
Interesting results. If I replace IJJ by MVV, I have MVV as the selection for each of the first three months,
with a total return similar to what you have - slightly better.
30 May 2014, 11:50 AM

Alexander, I dont think you have the signals quite right. In January we were in CASH, followed by SSO,
EDV, EDV, and ILF. June signal issued today is to stay in ILF. So far the system is up over 13% YTD.
See my blog post today re the modified GMR performance.

http://bit.ly/RL1QEz
30 May 2014, 11:56 AM
varan
If you want to go the Leveraged fund route, you might want to replace TLT/EDV by UBT. That will be a strategy on
steroids with the concomitant risks, many of them perhaps as yet unknown or at best unappreciated, of the
Leveraged ETFs. But the performance will be great as well.
30 May 2014, 12:06 PM

I just tried subbing UBT into my backtest and it lowered overall returns and the Sharpe ratio versus using
EDV.
30 May 2014, 12:42 PM
varan
What time period? It has been around only since 2011.
With my basket:
2011-2014 CAGR 20.5%
Replace IJJ by MVV, keep TLT

2011-2014 CAGR 33.1%
Replace IJJ by MVV, TLT by UBT
2011-2014 CAGR 37.8%
30 May 2014, 12:53 PM

You're using TLT. I use EDV. EDV is pretty volatile. As Grossman said, it behaves like a 1.5 leveraged
fund.
30 May 2014, 01:38 PM
varan
Not to belabor the subject, but to close this discussion:
MVV IEV EEM ILF EPP EDV
2011-2014 35.6%
MVV IEV EEM ILF EPP UBT
2011-2014 37.8%
30 May 2014, 01:59 PM
The attached two charts demonstrate the difference in strategy performance metrics for two exactly the same
strategies, apart from the time frame used.
- re-balancing at the close of each month
- momentum only, with (precisely) 3 months lookback
- re-allocate all capital into the best performing asset.
Backtests are done with AmiBroker, using synthetic ETF's created with R:
$EEM, $EFA, $EPP, $ILF, $MDY and $TLT.
Example: price history for $EEM = EEM + VEIEX + FEMKX.

(see here for a how-to: http://bit.ly/1m0uObO)
On the charts: a daily equity curve in red and a monthly equity curve in black.
Full backtest period 1993/12/31 - 2014/04/30:

http://bit.ly/1m0ujhV
Partial view (1999 - 2003):

http://bit.ly/1m0uhGP
With the same CAR of 19.81% (full 20+ years) cq 29.97% (last 10+ years), all other metrics are positively skewed
on the monthly time frame due to the (unrealistically) lower volatility.
So when comparing systems one might take the applied time frame in consideration too.
31 May 2014, 09:45 AM

May be helpful:
http://seekingalpha.co...
01 Jun 2014, 08:08 AM

That article struck me as pure rumination. There was no hypothesis and no conclusion. I am also
suspicious when it says "U.S. stocks crushed every momentum system we tested from 1995 to 2000 on a
cumulative annualized basis." I tested a momentum strategy that beat US stocks from 1995 to 2000 on a
cumulative annualized basis discussed here:
01 Jun 2014, 09:26 AM

It was in the context of the various "lookback" momentum strategies.
02 Jun 2014, 07:48 AM
designshoe
all the back testing isn't gonna give you any iron-clad formula to great returns going forward. the world will never
relive the past patterns again. that's the principle of uncertainty at work.
09 Jun 2014, 01:02 AM

Your comment is very constructive:
"all the back testing isn't gonna give you any iron-clad formula to great returns going forward." -- this is a
cliche. Of course we can't predict the future.
"the world will never relive the past patterns again" -- this is false. It happens all the time. Apparently, from
2003 to 2013, as shown in the back test, the world continued to follow the patterns shown by Mr.
Grossman.
"that's the principle of uncertainty at work." -- I'm not even sure what this means, other than a reiteration of
point 1 above.
Rather than baldly criticize something that someone has obviously put a lot of time and thought into, why
don't you try to apply the scientific method and seek to show some evidence that supports your views?
In any event, if you are correct then why invest at all?
09 Jun 2014, 02:06 AM
rabsparks
I really appreciate the time that everyone has spent on "pulling apart" GMR, et al. I too confess that I've
done my share of backtests on GMR, using Excel and Yahoo Finance.
But after spending many months emailing questions to Mr. Grossmann, talking to several of you via email,
and performing my own backtests, I decided to both subscribe to Mr. Grossmann's GMR and also keep
working at developing my own formula so that just in case anything happened to Mr. Grossmann, I could
still get a "pick" each month.
My constant testing was a bit tedious. I use three different data inputs for my monthly calcs. But I am
satisfied that I have a backup if need be.
That said, the current major issues are: although I have basically gotten the formula to the point that I
have faith in its picks, I have no idea how Mr. Grossmann picks his selection of ETFs other making certain
that they are not correlated. When I first subscribed to the GMR strategy, FG moved from IEZ to FEZ,
which obviously put the fear of God in me. Second the weighting (70/30) or whatever it is slips by me.
That said, it appears the at least for the time being GMR works. I know that it has worked for 2003 through
2013, but the future (like it or not) is still an unanswered question.
True, 2014 GMR started off with a flatulent sound but has seemed to straighten out and fly right. I have
two accounts. One is an IRA while the second one is a discretionary account. For these account I'm
showing a return of 2-3% per month and that puts a smile on my face.
I really tried to do the trades on the designated days. But February started off with a five day "No AC"
period which put a damper on my trading. The good news was that because I was kind of locked out in
January, my FEZ loss was turned into a profit (albeit minor) in spite of my intentions.
ETF Buy Date Sell Date ROI (Used Quicken for this report)
FEZ 11/1/13 2/11/14 0.46%

EDV 2/3/14 4/30/14 3.45%
ILF 5/1/14 to date 4.83%
EPP 4/3/14 5/29/14 2.60%
I learned not be in a rush to sell/buy the current picks. I am trying to trade within a few days of the first of
the month for both sells and buys.
I use Point & Figure charts to "watch" the selected ETF before I buy. Then I use the P&F charts before I
sell. I am not opposed to selling anytime in the month. For instance I am closely watching ILF and I have a
Limit Sell point ($38.93) just in case the bottom falls out of this highly volatile ETF.
I give Mr. Grossmann credit in that he has made one or two changes in GMR, adding EDV for a hedge if
EDV is Number 2 in his picks.
10 Jun 2014, 11:22 AM
varan
Except for EPP, that is the same as pure return based strategy for the basket of IJJ, FEZ, EEM, ILF, EPP
and EDV.
10 Jun 2014, 01:00 PM
designshoe
there is no evidence that any 10 year period repeats the previous 10 year.
you might have to go back 50-70 years. to detect similar set up as a zero interest rate policy, global debt deflation,
QE.... then again, there is no precedent for today's monetary policy since the removal of gold standard.
all it takes is one little blackswan event like 2008 "lehman", 2011 mini-crash, 2012's "grexit", 2014 "Ukraine" to
throw all the neat old patterns into disarray.
financial markets are driven by IRRATIONAL people

so it's a mass psychology that rules the price movement.
if you just want to be a noise trader and trying the catch the small waves, then keep trying to use this method of
market timing stocks/bonds. and just looking at ACCU or ICCIX perfomance, you'll easily see how well the pro's
have done trying to do momentum ETF trading. and they got PHD's with great algo's too.
there is a reason there is no billionaire market timer on the forbes richest list.
just think if you can get returns above 30% for 20 years, you'd be a billionaire already. but you aren't. and I'm not.
lot's of research has shown the best way to seek alpha to be a long term value investor in equities. especially
smallcap value... it's that's simple.
you decide what is value for yourself. nobody is gonna have a lock on that.
09 Jun 2014, 11:12 PM

"there is no evidence that any 10 year period repeats the previous 10 year."
Yes, there is ample evidence of such behavior. See the 20 year back-test below. The second ten years
continued with the same pattern as the first 10 years.
You can also look at the 23 year back test here:
You also state that there are no billionaire market timers. In fact there have been several successful
market timers. Today's article about technical analysis provides:
" The reality is that there are examples everywhere of successful investors applying technical and
momentum analysis to achieve strong performance -- one such example is famed and highly-successful
investor Richard Driehaus of Driehaus Capital Management. Another example is Cliff Asness of AQR."
It is frustrating to see Mr. Grossman's hard work criticized by unsupported arguments that are
contradicted by readily-accessible evidence.
10 Jun 2014, 02:47 PM
designshoe
good luck trying to be Driehaus or Asness.
again, they got PHD's and algo's you and I haven't got. and they still cannot beat the sp500 index over the 20 year
span. heck not even the great Bill Miller could. all it took was one "lehman" event to wipe out his sterling 13 year
record
11 Jun 2014, 01:31 AM
MCFA
Curious about the look-back period. I notice 60-70 days works better than 90 days. Anybody else get better
results with a 90-day look-back?
14 Jun 2014, 01:55 PM
MCFA, you need to separate calender days from trading days.
A typical year has some 252 trading days. This results in an average of 21 days per month.
So 3 months or 63 days puts us right in the range you mentioned.
90 days would mean a look-back period of 4 months and a couple of days
15 Jun 2014, 11:05 AM
doknabox
Frank, When do you find is the best time to do your trade switches? On the Open, around 10 am EST, around
3:30, on the close, or any other time? Which time do you think results in the least trading costs?
15 Jun 2014, 12:13 PM
varan
If the results depend on such minute details, the strategy is not worth it. That also applies to the number of
days in the look back period. Any result that you get by some sort of optimization process is very dubious
in its usefulness. The first rule of developing good robust strategies is to avoid optimization as much as
possible.
15 Jun 2014, 12:42 PM
berry
Very true and well stated and supported by extensive back-testing.
15 Jun 2014, 04:53 PM

Author’s reply » I normally do the trade within the first 2 days of the month. You don't have to time the
trade and you should not trade at market at the open of the first day of the month.
24 Jun 2014, 08:59 AM
doknabox
Thank you Frank for your response. It's been shown that the strongest days of the month are the first one
or two because of money flows. Why don't you invest in the last 2 or 3 days of the month to "beat the
crowd," so to speak?
25 Jun 2014, 10:25 AM
berry
He could do that, but that would be a different strategy than what he has promoted/backtested/sells
subscriptions for.
25 Jun 2014, 12:01 PM
doknabox
berry, what you say is true... but If the results depend on such minute details, the strategy is not worth it.
Any result that you get by some sort of optimization process is very dubious in its usefulness.
27 Jun 2014, 05:57 PM
berry
Would not trading two days before EOM to "beat the crowd" be "some sort of optimization process"?
I would imagine Frank has tested those kinds of alternatives/optimizat... and chosen the one he feels is
best overall, else why make the comments he has? We could just trade the strategy at random, or on
each ETF change, but that's not what he has proposed or is selling.
But, yes, I agree, based on my back-testing, the published results seem to be a bit too "fine tuned" for my
liking. Small changes in timing can have a big effect on overall results. But I do like and follow the strategy
in general. However, in back-testing, while changing the buy/sell date does change the results, I have not
found buying/selling on a different day of the trading month to yield repeatable improvements in
performance.
28 Jun 2014, 12:52 AM
Some time ago one of my blog visitors asked basically the same question. Since I was curious too, I
performed a little backtest. Please note the backtest was bond orientated, because my reader was
interested in Grossmann's BRS model (in that respect this post is slightly off-topic here).
Portfolio:
$CWB (CWB extended with VCVSX)
LQD
AGG
SHY
$EDV (EDV extended with TLT)
IEF
The $-sign indicates a synthetic ticker with extended price history using "R". See my blog for details:
http://bit.ly/1m0uObO
The backtest was performed over a 10 year period, starting end of May 2004. Re-balancing is based on 3
month performance (only), allocating all capital to the best performing fund (just like GMR).
The equity chart (link below) shows the equity curves for Offset_X, where X is ranging from 0 - 21 in steps
of 3.
At Offset = 0 the portfolio is rebalanced exactly at the end of the month.
At Offset = 3 rebalancing happens 3 trading days later and so on.
Offset = 6 had the worst performance and is not shown to prevent an even more cluttered graph.
The coloring labels guide you to the right curve, hopefully ;-)
Equity chart for Offset_X:

http://bit.ly/1pAGw4s
Table with system metrics for Offset_X (sorted on CAR):

http://bit.ly/1pAGw4u
For this particular lineup and backtest period, the best performance is with Offset_3 (red curve), closely
followed by Offset_0 (black red).
28 Jun 2014, 09:15 AM
berry
In my experience, the results are also dependent on the time period chosen - e.g. offset_3 may be best in
that particular period, but may not be in another. In these situations, it's impossible to speak in
"absolutes".
28 Jun 2014, 09:33 AM
doknabox
berry, thanks for your response. Optimization is said to be "the curse of trading" - well I don't agree. When
we choose our 6 ETFs to trade from the 4800 available, we're optimizing. When Frank picked EDV to
trade from IEI, TLH, TLT & EDV, he was optimizing. Designing these systems is nothing more than
optimization.
I think optimizing just hints at your future returns. I am very happy to get 50 - 75% of the optimized returns
(realizing that the TAA process will "hopefully" limit my drawdowns.)
I have communicated with some system developers/analyzers who's research demonstrates that trading
on certain days of the month gives one a small advantage.
I wonder what others think.
28 Jun 2014, 11:49 AM
QED, berry ;-)
Same backtest basics as the above, but this time with the GMR portfolio.
Equity chart for Offset_0,3,4,6,21:

http://bit.ly/1pBh1Aa
Table with system metrics for Offset_X (sorted on CAR):

http://bit.ly/1pBh1Ac
For this particular lineup and backtest period, the best performance is with Offset_4 (blue curve). Offset_3
(red curve) is second best and still 360% net profit behind. This time Offset_0 is fifth (black curve).
It depends indeed, but who was expecting anything else?
28 Jun 2014, 04:29 PM
varan
TLT and EDV are from the same asset class, both expected to be negatively correlated with the market,
with EDV effectively being a leveraged version of TLT.
So using EDV (or even UBT) almost suggests itself, but can be called optimization only in a legalistic
sense.
In any event, I think that (a) the use of optimization should be as minimal as possible, and not that no
optimization should be used at all, and (b) more you use optimization, less robust will be the resulting
strategy.
28 Jun 2014, 04:41 PM
berry
Yes. I've run tests across the entire monthly spectrum and over varying time periods and the differences
can be dramatic. And my back-tests confirm variations of double to triple results depending on the
monthly start date. If you look at the data in detail, you'll soon notice some very large single day moves for
some of the ETFs. Depending on timing, those large changes will impact the selection of etfs for a period
(or even longer) and can have a marked effect on overall results. For example, a very slight change in
timing changes a back-test for 2013 from 30%+ to just over 10%. I do not believe those variations are
"time/date" dependent, but rather the process is somewhat serendipitous - thus some good portion of the
performance variation is just luck-of-the-draw, and the selection of specific sell/buy dates is over fitting.
I personally believe "fitting" it to any specific buy/sell day is over-fitting. That said, there may well be some
"calendar month" variables that do come into play since in my tests, normally the eom/som strategy
performs as good or better than most.
28 Jun 2014, 05:17 PM
woody5023
I joined seeking alpha late last year. I was interested in learning more about TQQQ. I accidently stumbled upon
one of Frank's articles about his strategy. I've been a mutual fund investor since 1992. I was looking for a way to
take some side money and play aggressively in the market. Soon, I became fascinated with how SVXY worked. I
read through his strategies and locked on to his volatility strategy. I dont know why, but the more I read from other
sources and learned how they work, I felt more comfortable with it. So, I shared this knowledge with some other
people in my work center who also invest. Not a single one knew what ZIV was our anything about the VIX. They
were skeptical. I told them that i was doing it beginning the first of the year. So I subscribed. All the research said
you must have an exit strategy but not much is out there to say when and where to go when you exit. Frank does.
The information on his website that shows his strategy performance is accurate, however, what you may not know
is that each personal email contains a bit more information that you can use. For instance, when the Russia
Ukraine crisis hit in March, his strategy switched from ziv to EDV but in his message he warned us that this crisis
at the time is political and ate most likely short lived events. He gave us three options based on our own comfort
levels. I stuck with ZIV and haven't switched yet. I've even hedged with sorry TMV at one point.
My YTD performance is more than 17 percent with it so far but I have also used his information to take a long
stake in XIV and a short stake in TVIX. My overall return is 41% YTD.
I'm very happy with his information and since I'm invested more than 100k to start, you know I'ma happy customer
right now.
23 Jun 2014, 06:06 PM
pat12357
woody, when you say that your return is 41%, do you base this on the money that you've invested in these
strategies or is this the return of all your investments?
The problem that I have with these strategies (not Franck's in particular) is that I don't feel confident
enough that they are not overfitted in order to invest a large portion of my equity in any one of them.
24 Jun 2014, 09:15 AM
woody5023
My return with ZIV, XIV, TVIX and the TMV short position hedge have netted about a 41% return YTD so
far and growing. I took a 70k stake in ZIV (the strategy) to start. I took a 30K stake in XIV and a $50k
stake in TVIX after learning more about the VIX and how these things work. So, YTD I am up $14k with
ZIV, $14.5k with XIV, and $31.5k with TVIX.
I also have a couple other mutual funds in technology and dividend stocks.
I currently have long positions in PZZA, TSLA, TWTR, WPCS, and UVXY.
24 Jun 2014, 02:21 PM
pat12357
Thanks for the reply.
What is the risk you are exposing yourself to with ZIV and the likes?
25 Jun 2014, 12:31 PM
sarracenia21
First of all, I would like to say that Frank's articles have been very thought-provoking. The idea of using
momentum is catchy, and if it works, all the better.
What I worry a bit about is him understating the associated risks with a naked short of UVXY/long position
with XIV. In a situation where the VIX spikes suddenly, a person with substantial holdings of XIV or inverse
volatility in other forms will face substantial losses if not a total wipeout.
The reason I'm comfortable with trading ZIV is because large spikes in the VIX seem to most often occur
in semi-weak markets, in which a momentum strategy would have me out of inverse volatility. Plus, I don't
trade that much ZIV.
What CANNOT be discounted is still the probability of a one-day wipeout event. Even though most spikes
in the VIX occur in weaker markets where you will not be exposed, they CAN occur in healthy markets
and due to their inherent severe and unpredictable nature CAN put you out of business (a euphemism for
an irreparable amount of money).
By trading inverse volatility, you're being the insurance man. It is nice to sell flood insurance when
everyone's house is high and dry, but when everyone's houses are washed away, you will be holding the
bag. Right now, the reason you are having such massive gains is because everyone's house is fine.
However, eventually something will happen.
Woody, you seem to be an intelligent person, but I would urge a risk reassessment.
Last, my message is not to scare people away from inverse volatility. It is just a warning for those who are
not aware of its risks.
Cheers,
-S21
30 Jun 2014, 02:42 AM
DREW Seeker
Woody5023, you mentioned ZIV etc. but I was not aware that was part of Frank's Global Market Rotation (GMR)
strategy. Do you mind clarifying exactly which strategy you are using? Thanks, DREW
26 Jun 2014, 04:00 PM
berry
ZIV is an ETF in Mr Grossman's GMRE strategy.
You may view all of his strategies at his web-site.
26 Jun 2014, 04:16 PM
tsantoso
Hi Frank,
I am sure you received countless commendation on this great article and I wanted to thank you again. I just
wanted to clarify on your ranking methodology, as you mentioned on 10 Aug 2013 post:
“You can use the 20 day volatility averaged over 3 month.

For the ranking I calculate the 3 month performance of all ETF's and normalise between 0-1. The best will have 1.
Then I calculate the medium 3 month 20 day volatility and also normalize from 0-1.
Then I used Ranking= 0.7*performance +0.3*volatility.
This will give me a ranking from 0-1 from which I will take the best.”
You added the normalized momentum to volatility for your ranking. But wouldn’t it be more intuitive to do a minus
volatility (Ranking= 0.7*performance -0.3*volatility) or Ranking=Performance/Vo... As most Finance literature
would tell you volatility is not a desirable feature of an instrument and hence should play a part in reducing
(instead of increasing) the ranking of the etf?
If anyone has an opinion on this, feel free to comment.
07 Jul 2014, 09:52 AM
berry
Based on back-testing, Frank's method yields superior results.
While low volatility might appear to be more desirable, in testing, it is not.
cheers
07 Jul 2014, 10:40 AM
pat12357
If I remember correctly (it has been a while since I implemented it), Frank's model gives preference to
lower volatility. The volatility ranking is from low (top rank) to high (lowest), normalized.
07 Jul 2014, 11:08 PM
tsantoso
Hi pat and Berry,
So which one is the right one "+" or " -" volatilty *0.3
09 Jul 2014, 09:00 AM
berry
Apparently, only Frank knows for sure. Leastways, I do not.
He had mentioned too much volatility being bad, but my impression was "normal" was good.
09 Jul 2014, 09:08 AM

Berry, I had the same issue. I could not replicate Grossman's results with a negative coefficient for
volatility. Only a positive coefficient improved results but the system seemed less robust. In the end, I
abandoned using volatility at all in my own investing, which I seem to recall is a position you have
advocated in prior comments (forgive me if my memory is incorrect).
09 Jul 2014, 09:57 AM
pat12357
From the top of my head:
1. rank the funds by returns and calculate a return factor for each over the [0,1] interval (the top gets
ret_fac = 1, the bottom 0, etc.)
2. rank by volatility and calculate a vol. factor; here lowest volatility gets vol_fac=1, and highest vol_fac =
0.
3. final ranking is determined from: 0.7*ret_fac + 0.3*vol_fac
The above is based on the comments I read on SA. However, when I checked the rankings posted on
Frank's website, I saw that he is actually using a different formula (for the normalization). I don't remember
what it was, but it was not hard to reverse-engineer from the numbers. Nevertheless, it still ranks higher
funds with lower volatility (of course!).
09 Jul 2014, 11:13 AM
berry
Marc, you are correct regarding my earlier comments on volatility as a factor. My Back-tests varied
considerably depending of the time-period chosen. While I was unable to determine a consistent "top
performing" volatility component, my tests indicated that, for the majority of cases, a reasonable volatility
component, did improve performance and seldom hurt it so I'm currently using volatility as a positive
contributor in my ranking. I believe overall it helps more than it hurts.
In my tests, utilizing volatility as a negative component, hurt overall performance, but to each his own
That said, I believe a "performance only" strategy to be very sound also.
09 Jul 2014, 12:40 PM
tmdoherty
Hi tsantoso, pat, marc, and berry,
From a conceptual standpoint, it seems to me that the objective would be to normalize volatility. Trading
depends on volatility; if it is too low, that is perhaps just as bad as if it is too high. But there is another
reason: high-volatility ETFs might seem to demonstrate higher returns just because of the higher volatility
and the inherently greater noise that will necessarily accompany higher volatility. So a spike from noise
alone could spuriously influence returns temporarily and might not indicate a true change in relative
strength/momentum compared to the other ETFs in the basket.
I believe that from a computational standpoint, the way to normalize for variable volatilities is to either
calculate correction factors for each ETF and multiply returns by the correction factor, or divide by volatility
(as reflected by the SD of returns, or the SD of price movement). So, one could divide returns by the SD
of price movement over some specified time frame (say 3 or 4 months), or by the SD of returns over some
specified time frame. I believe these are equivalent approaches.
I also never understood Grossman's way to deal with volatility, and have never been able to replicate his
results. Cohen's approach just dispenses with attempts to adjust for volatility, and fortunately that seems
to work out just fine. But when you do that with Grossman's basket of ETFs, returns suffer significantly
(from around 44% per year to around 33% per year). On the other hand, overall volatility of returns (as
manifested in the SD of returns) decreases from 26% to 22%. But if you correct returns for SD of returns
(i.e., returns/SD), then that ratio decreases from 1.7 to 1.5.
So my interpretation is that risk-adjusted returns suffer by not correcting for volatility when using the
basket of ETFs that Grossman uses in his GMRS strategy (i.e., MDY, FEZ, EEM, ILF, EPP, and EDV, with
SHY used for a cash position). However, Cohn's version of the general Tactical Asset Allocation strategy
does not seem to suffer much if you don't correct for volatility. Cohn's basket of ETFs differs a bit from
Grossman's; Cohn uses SSO in place of MDY. This tells me that the selection of ETFs used in TAA can
make a big difference.
Terence
09 Jul 2014, 10:18 PM
pat12357
My advice:
- don't fiddle too much with numbers/parameters trying to improve performance; it leads to overfilling and the
testback results become meaningless
- base your strategy on a reasonable hypothesis (e.g. use a volatility factor to minimize whipsaw effect).
- test out of sample
09 Jul 2014, 11:28 AM
varan
FWIW I have not been able to understand the conceptual basis for the ranking based on volatility. Of
course you can make some assertion on how the inclusion of volatility in the rankings will effect the results
one way or the other, but I have not seen any convincing quantitative arguments.
In any case, since the basket of IJJ IEV EEM ILF EPP TLT based on three months returns alone yields a
CAGR of 28.8% (35% if IJJ is replaced by MVV)for 2003-todate, at least for me it does not make much
sense to complicate the strategy. Anyone should be happy to get 28% over a 11 year period.
09 Jul 2014, 12:13 PM
pat12357
Conceptually, one can imagine that a large spike in volatility during the look back period can result in false signals.
How important statistically this is, I have no idea. Especially when the basket contains funds with comparable
volatilities.
I have not examined carefully the volatility parameter but I am sure that if Frank is using it, it must improve the
back test performance of his strategy. This alone, however, is not sufficient justification for its use. I would start by
examining the data to see if the improvement is consistent or if it's a result of several isolated events that have no
statistical significance.
And I agree - 28.8% is outrageous. Complicating things always carries ignorance risk.
BTW, isn't risk parity based in similar concept - giving more weight to lower volatility?
09 Jul 2014, 12:47 PM
ikkyu
Frank uses a risk-parity style position sizing to use 30 year bond futures instead of EDV. Roughly you
need to lever to 1.5 to match the vola of EDV.
This is a huge advantage to trade these contracts over the thinly traded EDV.
10 Jul 2014, 12:12 AM
varan
Risk parity of course entails volatility minimization, although it does not use the return data at all.
I think that the idea that is closest to the mixed approach (i.e. using both returns and volatility) is the Kelly
fraction on the basis of continuous distributions, that leads to the weights being inversely proportional to
s/r^2 where s is the volatility (standard deviation of the returns) and r is the return. From that viewpoint,
the volatility term for ranking should have a negative sign. I don't know if that leads to better performance.
10 Jul 2014, 12:14 AM
ikkyu
Agreed. Volatility should be a negative ranking component.
Frank has carefully chosen funds with close to the same vola, so he is close to risk parity from the get go.
But the frisky bond part of the port is essential, requiring a strip 30 year fund (eg EDV) or leverage.
10 Jul 2014, 12:20 AM
ikkyu
On the volatility component... You all are much more mathematically sophisticated than me. That said, I use ETF
replay backtesting.
For the Global Market Rotation (2009-now):

-70% return/30% volatility yields
CAGR 30.4 Sharpe 1.25 Max DD -22.6
-for straight return

CAGR 23.7 Sharpe .92 Max DD -28.4
For Max Yield Rotation (2012-now):

-70/30 >>> weird it shows exactly the same return as the straight price performance. Might be a glitch.
-straight return
CAGR 55.6 Sharpe 1.98 Max DD -14.2
Cheers from Osaka,

john
10 Jul 2014, 12:03 AM
ikkyu
I showed many similar performance improvements adding volatility to momentum with various portfolios. I
wanted to ditch it as a parameter, but I have decided to keep it.
Frank seems to me to be a thoughtful guy who has considered these issues in detail. Of course we need
to do our own homework.
10 Jul 2014, 12:05 AM
varan
what basket?
10 Jul 2014, 12:36 AM
tmdoherty
I don't think it is fair to compare Grossman's MYR and GMRS performances. You've got apples and
oranges, because the MYR boosts its performance dramatically by shorting volatility whenever it can. I
think that's an artificial enhancement.
In contrast, GMRS does not short volatility.
MYR and GMRSE are frequently invested in ZIV, so they have become more of a hybrid between pure
TAA and volatility trading strategies. The MYR strategy only reports returns for the last 3 years. If you
compare the last 3 years returns for GMRS with MYR and GMRSE, they are:
GMRS: 45% (no volatility component)

MYR: 69% (short volatility)
GMRSE: 56% (short volatility)
I think it is apparent that reported returns on the latter two are substantially boosted by the inclusion of
ZIV. These two strategies are frequently in ZIV.
Nothing wrong with that really, but I just think one should be careful about comparing apples and oranges.
Terry
10 Jul 2014, 12:40 AM
varan
Those who are looking for optimal value of the parameters would do well to look at this recent paper:
http://bit.ly/TV05Wn
The details are a bit dense, especially for someone with only elementary background in probability and statistics,
but the basic ideas are quite accessible.
10 Jul 2014, 10:45 AM
GregTemesi
I agree. IMHO in-sample optimization and optimal values do not make much sense in this context, the
published backtested results (CAGR etc.) of the strategy are clearly overfitted, confidence interval is too
large. Frank is also using walk forward optimization (WFO) and the whole point of the strategy is that it is
capable of exploiting the momentum anomaly also with WFO (I implemented different variants of the
strategy in Ninjascript). An other great paper on the topic:
http://bit.ly/1mwv4o5
14 Jul 2014, 02:09 PM
EdwardjK
I have been using Mr. Grossmann's Bond Rotation Strategy since February, 2014. As part of my due-diligence, I
capture the daily data used in his BRS strategy and attempt to validate his monthly allocation.
In various posts, Mr. Grossmann says that his strategy uses a combination of 70% return and 30% volatility to
determine each asset's ranking. That is directionally correct.
Using Excel, I can confirm, at least for the BRS strategy, that the 70%-30% is adjusted each month. And the two
are ADDED together, not subtracted. For example, in the May, 2014 ranking, Grossmann used a 75%-25% (going
by memory here) allocation. I matched his ranking values exactly.
Interestingly, using 70%-30% would not have changed the ranking, although the rank value of each asset would
have changed.
I have written Mr. Grossmann several times on this point and he has declined to describe the technique he uses to
determine the split between return and volatility.
I imagine he does the same for all his other strategies.
17 Jul 2014, 04:15 PM
AlexanderHorn
We've started an Instablog recently including the GMR, see the 2014 Year End summary and some 2015 Portfolio
Ideas there:
http://bit.ly/14e4rNZ
http://bit.ly/1smcYcA
12 Jan 2015, 04:50 PM
thjames
How do you normalize the returns and vol? is it simply the percent rank for each individual data set? So over a 10
period sample set the lowest vol gets a 1 and the highest a 0? or is it more complicated than that. If it is anyone
can point me to some research/reading to enlighten me? thanks
20 Feb 2015, 10:34 PM
AlexanderHorn
You can either use Z-Score or Scaling into range [0-1]:
Z-Score of a raw score x [1] is
z = {(x- μ) / σ}
where:
μ is the mean of the population;

σ is the standard deviation of the population.
Scaling is easier, like X’ = (X – Xmin) / (Xmax – Xmin)
21 Feb 2015, 09:17 AM

I also could not duplicate these results using the plain 3 month look back. One problem is that the EEM fund did
not commence trading til April 2003. Therefore the results for that year are sketchy. I also get different monthly
holdings for that table at the end of article.
http://bit.ly/1PrtEtq
13 Nov 2015, 08:27 AM
drftr
Frank,
You may want to re-visit your original strategy after reading Trendxplorer's article on PAA (Protective Asset
Allocation). He just posted results using PAA combined with Varan's "Simple GMR" which is based on your
strategy and by doing so was able to get the same returns BUT improving MaxDD from -37% to a fantastic -17%.
What's not to like!?
drftr
15 May 2016, 04:51 AM
Rensys Stocks
The simple 3 month rotation strategy does not work any more from mid 2014 on. Seems like incredibly over
optimized although simple logic.
23 Jun 2016, 04:42 AM

Author’s reply » You are right, GMRS is at the moment our strategy with the lowest YTD return. Reason is that it
is a global strategy and such was hit hard by the strong rise of the US$ which had a very bad effect on unhedged
foreign market ETFs. However 1.2% YTD is still slightly better than the All World Country Index.
In general because high currency volatility, it was better to remain in US$ nominated ETF strategies. These did
very well even YTD 2016 with performances up to 21%. So the methodology itself is working.
See performance table on our website: http://tinyurl.com/zau...
Strategy/symbol year to date % ▴

UIS-SPXL-TMF 21.59
MYRS 20.31
WORLD-TOP 4 8.73
GOLD-USD 7.47
BRS 7.38
UIS 7.08
GSRLV 6.34
BUGLEV 6.21
NASDAQ 100 5.59
AGG 4.23
BUGST 2.67
SPY 2.60
GMRS 1.22
23 Jun 2016, 09:01 AM
tmdoherty
@Frank,
C'mon. That is total nonsense, and you know it. You should be ashamed.
For starters, you are making after-the-fact excuses. Worse, you are trumpeting all but meaningless YTD
performance while neatly stepping over the remnants of your flagship GMR strategy, that lost lots of money for
your subscribers.
Please explain what you think the interpretation of 6 whole months worth of data might be. Are you suggesting
that based on 6-months' data, people should subscribe to your service and invest money in your strategies? Is
that what you call due diligence? Would you invest major amounts of YOUR money in any strategy based on 6
months' performance data?
Let's instead look at data since live trading began. Your own charts show that the minute live trading began, the
strategy began to fail. It has gone down ever since, and currently shows a return from that date of -4.92%. In
contrast, again according to the chart on your website, the benchmark (VGHEX) has gained 6.32%. Now, the
benchmark is invested in over 800 stocks from more than 20 countries, and is thus exposed to the same currency
and country risks that GMR is exposed to, but has done far better.
(as an aside, it is not easy to understand how an equity mutual fund serves as a good benchmark for a strategy
that invests in both stocks and bonds of various sorts, but that's a relatively trivial issue)
GMR was never described or marketed as an investment strategy that attempts to identify, properly interpret, and
capitalize on macroeconomic factors like the relative strengths or weaknesses of currencies.
GMR was marketed as an algorithmic strategy that was agnostic as far as market direction or market forces was
concerned because it would simply follow momentum. Hence, it would do well in pretty much any market
environment. That has not been the case for several years, and prior to that, it was only the case with backtesting
that was obviously optimized. Now, once again, we see what happens when backtested and optimized strategies
are traded with real money prospectively.
I know, I know: now you've eliminated the flaw by incorporating currency hedging. But this is just further curve
fitting, so you are trying to solve the problem with the very same approach that caused the problem in the first
place.
Admit it: the GMR strategy is fundamentally flawed. If that strategy is flawed, how can anybody be sure your other
strategies are not similarly flawed for exactly the same reasons? Will readers be posting similar complaints to you
2 years from now about your strategies listed above?
Very likely the question is not IF that will happen, it is WHEN. There are many reasons for this, but one obvious
one aside from the fundamental issue of curve fitting is that your testing is inordinately skewed by the massive bull
market that we've seen since 2003 (with a relatively short bear market in 2008-2009). We and many others have
found that such a limited testing period is unacceptable because it involves only very limited market environments
that are not sufficiently reflective of what investors can anticipate going forward. 13 years testing is simply not
sufficient, and that doesn't even begin to get at the issue of out-of-sample testing (or rather, lack thereof, in this
case).
This may seem like overly-harsh criticism, but you are selling a product here. If you were just posting things for
free with no inherent conflict of interest, that would be a much different story. There, I would simply disagree with
you, and explain why.
But that's not the case. You post here to try to sell your products to new, unwitting customers. So it is only fair that
those potential new customers---typically and unfortunately somewhat naive, if unwittingly so---be forewarned
before they risk their money and pay you a monthly fee.
Furthermore, potential customers: there is nothing that Grossman offers that cannot be obtained absolutely free.
And, in fact, you can access far better strategies, again absolutely free, right here on SA.
TMD
07 Jul 2016, 08:07 PM
mohk1234
Just curious on a few other strategies that you have found on SA for free. Thanks in advance!
15 Jul 2017, 12:34 AM
IndyDoc1
@ TMD
For Franks's credit, the original GMR and bond rotation strategies were explained in detail in his articles, including
the fund selection, look back period and frequency of trading. The bond rotation strategy ( which in my opinion the
best that Frank came up with ) has been trading out of sample for over 2 years now; not a long time but so far, it
lived to its promise. Cliff Smith's low volatility trading strategy has independently reached the same concept as
frank 's bond rotation strategy
07 Jul 2016, 11:58 PM

Author’s reply » We´ve just published our Portfolio Allocations for 2017 employing this strategy, have a look here:
03 Jan 2017, 12:15 AM

Author’s reply » Our strategies are now available as Segregated Managed Account through The Estate Planners
Group: http://bit.ly/2lmUVSj
07 Feb 2017, 01:44 PM

A Global Market Rotation Strategy With An Annual Performance of 41.4% Since 2003 - Seeking Alpha

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

A Global Market Rotation Strategy With An Annual Performance of 41.4% Since 2003 - Seeking Alpha

Hochgeladen von

Copyright:

Verfügbare Formate

9/7/2018 A Global Market Rotation Strategy With An Annual Performance Of 41.

4% Since 2003 | Seeking Alpha

A Global Market Rotation Strategy With An Annual Performance Of

Sounds too good to be true! My first contribution on Seeking Alpha.

The Global Market Rotation Strategy (NYSE:GMR)

Annual performance (OTCPK:CAGR) = 41.4% (S&P 500=8.4%)

These global markets and ETFs are:

US Market (MDY- S&P MidCap 400 SPDRs)

During market corrections I invest in:

US Treasury Bonds (EDV- Vanguard Extended Duration Tsy (25+yr))

Return to Risk ratio (Sharpe ratio)

Strategy (investment 2003 - today) Sharpe ratio Volatility Annual return

SPY S&P 500 investment 0.38 20.3% 8.6%

5 Global Markets alone 0.73 28% 21.3%

5 Global Markets + Cash exit 1.25 23.3% 33.8%

5 Global Markets + Treasury 1.38 25.6% 41.4%

AGG - iShares Core Total US Bond (4- 0.40 5.4% 4.3%

Annual performance of the Global Market Rotation Strategy

Year GMR performance S&P 500 performance

2003 34.9% 28.2%

2004 24.6% 10.7%

2005 36.8% 4.8%

2006 39.4% 15.8%

2007 38.5% 5.1%

2008 55.4% -36.8%

2009 56.7% 26.4%

2010 40.0% 15.1%

2011 89.4% 1.9%

2012 14.2% 16%

2013 until end July 19.4% 21.2%

How to rotate the ETFs

Here is an example of the investments since September 2012

Date Portfolio GMR S&P 500

month ETF monthly return monthly return

Sep.2012 EPP 2.8% 2.5%

Oct.2012 EPP 2.5% -1.8%

Nov.2012 EPP 1.8% 0.6%

Dec.2012 EPP 3.8% 0.9%

Jan.2013 EPP 3.5% 5.1%

Feb.2013 MDY 0.8% 1.3%

Mar.2013 MDY 4.8% 3.8%

Apr.2013 MDY 0.6% 1.9%

May.2013 MDY 2.3% 2.4%

Jun.2013 MDY -2.3% -1.3%

Jul.2013 MDY 6.7% 4.8%

I would not suggest the optimization as that might lead to overfitting.

I use a script in Mathnium.

09 Aug 2013, 10:54 AM

Frank Grossmann, Contributor

10 Aug 2013, 12:18 PM

10 Aug 2013, 06:19 PM

31 Oct 2013, 12:51 PM

Kevin Wilbur, Contributor

Thanks for sharing, Frank. And welcome to the community.

09 Aug 2013, 11:14 AM

Frank Grossmann, Contributor

10 Aug 2013, 12:18 PM

There are two points I am not sure if I understand correctly.

09 Aug 2013, 11:23 AM

Frank Grossmann, Contributor

10 Aug 2013, 12:18 PM

10 Aug 2013, 01:55 PM

12 Aug 2013, 12:43 PM