Seeking Alpha

varan's  Instablog

Send Message
Individual investor.
  • A low turnover high performance strategy for trading ETFs 27 comments
    Jan 15, 2012 4:47 PM | about stocks: SPY, TLT, XLE, XLK, XHB
    A strategy for trading ETFs about four times a year appears to provide decent returns in backtesting for the years 2001-2011.

    The universe of ETFs that was used for back testing the strategy was taken from here

    but all the ETFs containing the words VIX or Futures in their descriptions and the ETFs whose dollar volume was less than $90M (as of 1/15/12) or more than $1b according to the site mentioned above were removed from the list. This process leaves a total of 51 ETFs in the basket.

    The following strategy works quite well in backtesting.
    1. Start from the close of the first full week of the year.
    2. Rank the performance of the funds during the period of 8 weeks ending on the close of the previous week.
    3. Invest in the top ranked ETF for 15 weeks.
    4. Repeat steps 1 to 3 for the whole year, closing out the position in the last week of the year.
    5. Repeat the steps 1 to 4 for every year.

    Back testing yields the following results for the period 2001-2011 (All results are based upon the weekly adjusted close data from I will be glad to collaborate if a reader wants to repeat the calculations with data that he thinks to be more reliable.):
    • CAGR 18.5%
    • Number of years of losses: 1 (out of 11, in 2002)
    • Maximum loss: -23% (in 2002).
    • Standard Deviation of Annual Returns: 22%
    • Square root of 'semi-variance': 14% (uses only those annual returns that are less than the mean return ).
    • Mean of Annual Returns: 20%
    • 1, 3, 5, and 10 Year CAGRs: 19%, 34%, 29%, 19%
    • 2011 Return: 19%
    • 2008 Return: 28% (mostly due to positions in SH, entries into which were made using the mehanical ranking strategy and no other consideration).
    All the trades for this strategy are available in the following spreadsheet, the first row of which contains the list of all ETFs used.

    For the first 15 weeks of the year 2012, the strategy is invested in XHB.

    In case one is concerned about relying on a single ETF traded every 15 weeks, we list here the results for the case in which investment is made in four top ranked ETFs every 15 weeks. This 'moderate' strategy leads to the following results:

    • CAGR 15%
    • Number of years of losses: 2 (out of 11, in 2002 and 2008)
    • Maximum loss: -19% (in 2002).
    • Standard Deviation of Annual Returns: 16.5%
    • Square root of 'semi-variance': 13% (uses only those annual returns that are less than the mean return ).
    • Mean of Annual Returns: 16.5%
    • 1, 3, 5, and 10 Year CAGRs: 6%, 19%, 17%, 15%
    • 2011 Return: 6%
    • 2008 Return: -6%.
    Although there is some improvement in volatiity, its not clear if it compensates for the quite substantial loss of CAGR.

    For the first 15 weeks of 2012, this 'moderate' strategy is invested in XHB, XLP, XLV, and TLT.

    This is not investment advice, but just an effort to track the results of hypothetical investments according to a particular strategy.

    I have positions in some of the ETFs in the basket of ETFs for which the back tetsing was performed.
Back To varan's Instablog HomePage »

Instablogs are blogs which are instantly set up and networked within the Seeking Alpha community. Instablog posts are not selected, edited or screened by Seeking Alpha editors, in contrast to contributors' articles.

Comments (27)
Track new comments
  • Fredrik Arnold
    , contributor
    Comments (1711) | Send Message
    Hi varan!
    Thanks for your calculations. I have three questions. (1) Did you use the (a) market order strategy or (b) the trend analysis methodology with stops to make your hypothetical investments in selected ETFs & ETNs? (2) How does one make 15 week segments within a one year period? Thirteen weeks works evenly. Perhaps the investor takes a break for seven weeks at the end of the year to decide on the next move. (3) Did you account for newly created etfs coming into the available pool each year for selection or did it matter?
    Regards, FA
    16 Jan 2012, 08:51 AM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » 1. Orders are assumed to be placed to be executed at the close on Friday.


    2. Somehow 15 weeks interval, with the last period obviously less than that works better than 4 thirteen weeks. One could just continue on into the successive year, but I find this way to be more convenient.


    3. I use all the ETFs available at a given time for which the data of prior eight weeks is available. So the newly created ETFs are used whenever they are thus available.
    16 Jan 2012, 10:52 AM Reply Like
  • Market Map
    , contributor
    Comments (200) | Send Message
    Looks like 50% of the tradesare "country" funds . I wonder what the results would be if you stripped those out and just stayed "domestic" .. ?
    16 Jan 2012, 10:09 AM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » I don't know if the other etfs are purely domestic.


    In any case, why bother?
    16 Jan 2012, 02:51 PM Reply Like
  • tmcdonough10
    , contributor
    Comments (2) | Send Message
    I think the question was asked earlier but I don't think it was given enough consideration, so I'll rephrase and pose it again: Did you account for survivorship bias in your backtesting results? That is, if you just take ETFs that are around today and run your tests back in time, you will be missing all of the etfs that failed in the past that might otherwise have decreased CAGR. I know in Faber's book he specifically mentions going back in time to get all of the data to account for this, so it is possible, and I was just wondering if you'd done the same. Thank you!
    9 May 2012, 09:44 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » I agree that there is a survivorship bias in this approach. It will require a lot more work to remove that bias since the historical data for the dead ETFs is very difficult to find (I have used data from Yahoo Finance, which quickly stops providing data for ETFs that cease to exist).


    Another problem in working with ETFs is the short length of the period for which these instruments have been available.


    In retrospect I should have emphasized both of these concerns in the blog.


    Thanks for your interest.
    9 May 2012, 11:04 PM Reply Like
  • tmcdonough10
    , contributor
    Comments (2) | Send Message
    No problem, thank you for all of your work. I was wondering, giving these concerns, do you still believe your system to be superior to a simple 5 sector momentum rotation model (e.g. pick the top 1/2/3 of domestic stock, foreign stock, commodities, bonds, and real estate based on avg of 3/6/12 month returns and rotate monthly) as posed in Faber's book, or do you think the CAGR/volatility will change materially if survivorship bias is accounted for? If so, why? I'm trying to set up a portfolio of my own, and am really interested in your system. If there's additional work you'd recommend I'd do instead of just feeding me answers, I'm all ears as well.
    10 May 2012, 02:43 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » I view this just an adjunct to more comprehensive strategy. A singe ETF even if updated periodically cannot constitute a complete portfolio.


    I think that starting with a basket of stocks with both equity ETFS and bond ETFs of various styles (for example IYY, IJS, IYR, IYE, EFA, EEM for equities, TLT, IEF, AGG, EMB. MUB amd MBB for bonds) and selecting 5 to 6 of them every 13 or 26 weeks, or even annually, on the basis of their relative performance in the prior few weeks is a strategy worth further investigation.
    10 May 2012, 11:59 PM Reply Like
  • MrBobDobalina
    , contributor
    Comments (70) | Send Message
    Varan - just getting a chance to test this. I thought I'd give it a try. It's one thing I can easily test (I used Using the exact etf's you suggested, the best combination (in my mind) is selecting the top3 each quarter based on the prior 2 months return (no MA filter).


    Results: 2003-Dec2012 (in earlier years, only those eft's available were considered and new etf's were added as they became available).


    CAGR: 16.1%
    MaxDD: -15.9%
    Sharpe: 0.84
    72.5% of all quarters were positive
    worst year - 0% in 2010.


    Not bad. Other combinations worked ok, too. But for me, id you can get a CAGR greater than the maxDD, you have something pretty good - or at least worth exploring. This was the only one that met that criteria.
    26 Dec 2012, 02:59 PM Reply Like
  • littletiger101
    , contributor
    Comments (10) | Send Message
    any update (july 1st) on the portfolio rebalacing under this strategy?
    1 Jul 2012, 12:11 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » Thanks for your interest.


    It is to be updated on July 30.


    So far YTD.


    1/6/2012 to 4/20/2012 XHB 14.55%
    4/20/2012 todate XLK -1.51%


    YTD Return 12.82%


    YTD Market
    QQQ 15.37%
    SPY 9.48%
    DIA 6.70%


    Average QQQ,SPY, DIA YTD: 10.52%
    1 Jul 2012, 12:31 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » Update:


    YTD Return 14.4%


    XHB 1/6/2012 thru 4/20/2012 14.55%
    XLK 4/20/2012 to date -.14%


    Update according to this strategy:


    At close on 8/3/2012 Buy signal for EWW (based on the performance during the period 6/1/2012 thru 7/27/2012).
    29 Jul 2012, 04:34 PM Reply Like
  • owldawg
    , contributor
    Comments (3) | Send Message
    Hi varan,


    This looks very interesting. My background is technical (researched fuzzy logic pattern recognition methods for 30 years) but I know very little about data driven investment strategies. Can you, or any other reader, give a little basic information on any of the following? Thanks.


    1 Is there a good book or other source of background for this type of investing methodology?


    2 What is the most convenient way to get the data needed for these strategies? (free, I hope)


    3 What is the most convenient technical computing environment to use for this work? Are you using Excel or some such general tool, or are you using tools inside a particular trading platform?


    I apologize for being a parasite and not contributing something interesting, at least not yet.


    Thanks, in advance.


    25 Sep 2012, 06:12 AM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » I welcome comments from the readers.


    1. Mostly from the web. Google tactical allocation. There are lots of papers that you can find here as well.



    2. I get my data from Yahoo finance.


    3. I have built my own set of tools within a Matlab like technical computing environment.
    25 Sep 2012, 10:01 AM Reply Like
  • owldawg
    , contributor
    Comments (3) | Send Message
    Thank you. This is plenty enough to get started.
    25 Sep 2012, 12:03 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » By the way, for downloading the data I use the httpclient library from Apache ( ) within the 'technical computing environment' of Mathnium ( ) . Someday I will release the library of functions that I have built to do the computations whose result I present here from time to time.
    25 Sep 2012, 12:10 PM Reply Like
  • owldawg
    , contributor
    Comments (3) | Send Message
    Thanks again. I like MATLAB and will use it or maybe Mathnium (which I did not know about) for the number crunching. Thanks also for the additional info on getting the data, which for me will be the trickiest part.
    25 Sep 2012, 09:56 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » Update 10/1/2012


    YTD trades


    XHB 1/6/2012 thru 4/20/2012 14.5%
    XLK 4/20/2012 thru 8/3/2012 1.27%
    EWW 8/3/2012 holding 3.32%


    YTD 19.8%
    Average of SPY/DIA/QQQ 17.4%
    60/40 60%SPY 40% TLT 11.6%
    Equally Weighted SPY, DIA, QQQ, LQD, TLT 13.3%


    Next update: 11/16/2012
    1 Oct 2012, 03:06 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » The new buy for 11/16/2012 is TLT. The next best performer is SH (inverse SPY). So this system is reacting to the current market downturn. (It's noteworthy that last year around this time, the things had started to turn around, and XHB was picked by the system, yielding 7% at close of the year.)
    15 Nov 2012, 02:52 AM Reply Like
  • Lieuallen
    , contributor
    Comments (6) | Send Message
    I'm very intrigued by this strategy, and would like to further understand its rationale and design. Specifically, I'm wondering why you decided to make it start at the end of the first full week of the year, and why you "skip" the current week when establishing the 8-week ROC.


    I'd also like to assist (if I can) with the evaluation of this system. I have implemented it in Fidelity's Wealth-Lab, using daily price data. My system replicates your results, with the single exception of ranking VNQ as the high performer for the last quarter of 2009.


    I've run a few different variations (buying at the open instead of the close on Friday, and then trying buying at the open on other days of the week). I've also modeled evaluating the 8-day ROC on different days of the week, with some significant differences in the results. This is a bit troubling to me, as it suggests that the system is very fragile and may over-optimized for past data.


    Any interest in a collaboration?
    6 Dec 2012, 06:13 AM Reply Like
  • Zach Tripp
    , contributor
    Comments (390) | Send Message
    Thank you for keeping this up to date Varan, nice work.
    20 Dec 2012, 09:42 PM Reply Like
  • wesfrank
    , contributor
    Comments (2) | Send Message
    I have starting trading this strategy recently and was curious if you were able to see if it was prudent to short the worst performing ETF based on this same screen? Thanks for your work, been following your posts for quite some time now.
    10 May 2013, 10:12 AM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » Thanks.


    I have not considered shorting.
    10 May 2013, 11:50 AM Reply Like
  • mpd3892
    , contributor
    Comments (14) | Send Message
    Any updates for 2013? Any changes to the formula? I'm looking for a something similar to apply to a 529 Plan. Any guidance on an effective once a year system that would enhance returns within a 529 Plan?


    As always, your ideas are much appreciated and have helped make me a more knowledgeable investor. In any event, thank you for sharing and if you ever decide to start your own site, please make it known.
    11 Jul 2013, 11:08 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » Thanks.


    This year the strategy is broken, although the long term record is still quite good.


    2000 6.53%
    2001 20.24%
    2002 -23.16%
    2003 47.87%
    2004 12.26%
    2005 11.74%
    2006 16.38%
    2007 17.82%
    2008 28.52%
    2009 63.31%
    2010 24.27%
    2011 19.84%
    2012 10.91%
    2013 -8.66%
    ANNUAL 16.54%


    For 2013, here are the details:
    RSX 1/ 7/2013 TO 4/22/2013 -11.66%
    EWJ 4/22/2013 TO 7/ 8/2013 3.39%
    2013 -8.66%


    The simplest once a year system that I have seen in this:


    On the first trading day of every year rank the performance of each of VTSMX, IJJ, IJS, VUSTX, VBMFX, and FNMIX during the last quarter of the previous year, and invest in the equally weighted portfolio of the four that did the best.


    2001 8.35%
    2002 -6.85%
    2003 32.94%
    2004 15.03%
    2005 11.07%
    2006 9.87%
    2007 0.33%
    2008 -7.41%
    2009 17.07%
    2010 18.20%
    2011 1.20%
    2012 19.06%
    2013 10.89%


    CAGR 9.82%


    The all ETF version consists of VTI, IJJ, IJS, TLT, AGG and EMB, and yields the following results:


    2008 -8.14%
    2009 6.05%
    2010 18.39%
    2011 1.22%
    2012 18.56%
    2013 11.30%


    CAGR 8.13%


    The performance of the two versions is quite different for 2009, and is not unexpected, as EMB and FNMIX are not exactly the same.


    All the caveats apply: past performance is not a guarantee of future returns, I am not a financial adviser, etc.
    12 Jul 2013, 01:51 AM Reply Like
  • wesfrank
    , contributor
    Comments (2) | Send Message


    I started trading this strategy on 4/29/13 and based on your parameters I found that IBB was the top performing ETF based on the "lookback" period of 2/22-4/19, it performed ~6% better than EWJ. Just curious why the difference in the selection of the top performer?


    Thanks for any info, so far I've done pretty well in the short time frame of using this strategy and you Simple Polygamous Strategy (I've used that one for a little longer timeframe).




    15 Jul 2013, 04:45 PM Reply Like
  • varan
    , contributor
    Comments (3514) | Send Message
    Author’s reply » OK I looked at the basket that I used and it did not have IBB, which indeed shows up if you add it.


    Looks like you started exactly at the right time.
    15 Jul 2013, 08:30 PM Reply Like
Full index of posts »
Latest Followers


More »

Latest Comments

Instablogs are Seeking Alpha's free blogging platform customized for finance, with instant set up and exposure to millions of readers interested in the financial markets. Publish your own instablog in minutes.