varan's  Instablog

varan
Send Message
Individual investor.
  • Simple GMR 304 comments
    Jul 31, 2014 10:31 PM | about stocks: EEM, IJJ, ILF, EPP, TLT, MVV

    Starting in August 2014, I will be tracking the results of the simplified Global Market Rotation strategy applied to baskets of ETFs. On the first trading day of every month, the strategy invests in the asset that performed the best during the immediately preceding three months on the basis of total return (i.e. including dividends and any other distributions). The back test results for three baskets (the mutual fund basket has been added mainly to get an estimate for the returns for a longer time period) for the periods ending on 7/31/2014 follow:

    RS-GMR-ETF: IJJ, IEV, ILF, EPP, EEM, TLT

    RS-GMR-LETF: MVV, IEV, ILF, EPP, EEM, TLT

    RS-GMR-MF: FDVLX, FIEUX, FEMKX, FLATX, FPBFX, VUSTX

     

    Period

    CAGR

    Sharpe (Sortino)

    Max. Drawdown

    Min. Annual Return

    RS-GMR-ETF

    2003-2014

    28.6%

    1.3 (2.8)

    17.2%

    6.5%

    RS-GMR-LETF

    2007-2014

    31.5%

    1.12 (2.11)

    22.4%

    4.1%

    RS-GMR-MF

    1991-2014

    20.7%

    0.97 (1.93)

    24.6%

    -24.6%

    YTD Returns

    RS-GMR-ETF 13.1%

    RS-GMR-LETF 4.1%

    RS-GMR-MF 14.1%

    For August 2014, both of the ETF strategies are going to be invested in EEM.

    The following figures display some results for the RS-GMR-ETF and the RS-GMR-MF baskets.

    (click to enlarge)

    (click to enlarge)

    (click to enlarge)

    (click to enlarge)

    Disclosure: The author is long EEM.

    Additional disclosure: This is not investment advice in any form.

    Stocks: EEM, IJJ, ILF, EPP, TLT, MVV
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 (304)
Track new comments
  • Stanford Chemist
    , contributor
    Comments (1974) | Send Message
     
    Interesting. .will follow!
    31 Jul 2014, 10:52 PM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    I agree with the Chemist. Sounds cool.
    1 Aug 2014, 03:24 AM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    Thanks, Varan.
    1 Aug 2014, 08:01 AM Reply Like
  • Steadyprofit
    , contributor
    Comments (42) | Send Message
     
    Me too, will follow. Have been tracking other GMR articles and discussions and it is a subject I follow with a lot of interest. Thanks!
    1 Aug 2014, 10:23 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    I have been doing similar strategy but with more funds. I look forward to your updates!
    1 Aug 2014, 10:29 AM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    Looks very interesting.

     

    Can you briefly explain the thinking behind the difference between RS-GMR-ETF and RS-GMR-LETF (IJJ vs. MVV)?

     

    Is that 4.1% YTD for RS-GMR-LETF a typo? (14.1%?)
    1 Aug 2014, 10:32 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @Left Banker

     

    The result is somewhat surprising, but it is not a typo. The LETF was invested in MVV last month. MVV lost over 10% in July. ETF was invested in EEM, which was barely above water, but did not lead to a loss.
    1 Aug 2014, 10:41 AM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    Varan,

     

    Do you think that that the lower CAGR for the mutual fund version of this strategy simply reflects a longer backtest period, or is there some inherent quality of the mutual funds you use that might explain it?
    1 Aug 2014, 11:06 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » My first instinct is to say that 28% CAGR for the ETFs is too high to persist for a long time, and so the mutual fund results are probably more realistic, even though still pretty good. In this regard my baseline is BRK-A: any strategy, that can yield a CAGR of around 15% or higher during 1991-to-date with much less volatility (as determined by Sharpe being significantly higher than the Sharpe of 0.65 for the Buffett's fund) and with very little degradation of the performance during 2000-to-date, is probably as good as one can normally expect.

     

    There may be some influence of the idiosyncratic risk of the mutual funds too, as they are all purported to be actively managed.
    1 Aug 2014, 12:01 PM Reply Like
  • spielerman
    , contributor
    Comments (280) | Send Message
     
    Thanks Varan,
    A couple of questions here as I've been trying to balance out the various strategies and methodologies over time periods as well as different investment accounts.

     

    -in my 401k I am stuck with the 91 trading day window on fees with switching funds. Does this strategy work as well on a 90 adjustment instead of monthly?

     

    -any idea on tax efficiency, or this is a matter of pay your taxes because the absolute returns, risk adjusted returns, and max drawdowns means.... success?
    1 Aug 2014, 12:24 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » For three month update period the maximum drawdown and maximum annual loss are too high.

     

    ETF (2003-2014)

     

    CAGR 22% Sharpe 1 Sortino 2 Max Drawdown 35% Min annual return -13%

     

    Mutual Funds (1991-2014)

     

    CAGR 15.7% Sharpe .73 Sortino 1.3 Max Drawdown 46.4% Min annual return -41%

     

    I am afraid that I am not qualified to make any statements about tax efficiency.
    1 Aug 2014, 12:39 PM Reply Like
  • retring investor canada
    , contributor
    Comments (18) | Send Message
     
    I am curious as to why you opted for only one leveraged ETF (NYSEARCA:MVV) in the portfolio RS-GMR-LETF, when there are leveraged versions of EEM, TLT, ILJ and EPP.
    2 Aug 2014, 08:22 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Excellent question. The main reason is that not much historical data is available for many leveraged ETFs: for EMLB and UBT only since 2011, for URR since 2009, etc.

     

    For the limited dataset, the back testing suggests that using UBT instead of TLT (and not URR for IEV or EMLB for EEM) provides the best possibility of improvement for the leveraged version. The CAGR for 2011-2014 with UBT and MVV increases from 27% to 36%, and more importantly the YTD return improves from about 4% to 20% (reflecting the fact that TLT has done well this year). Addition substitutions by URR and EMLB actually degrade the performance.
    2 Aug 2014, 02:30 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    According to my calculations, here's what you get when you substitute UBT or TMF (3x leveraged) for TLT for the time period Feb 2010 thru the present:

     

    With TLT:
    Win Rate: 69.2%
    Median Rolling 12-Mo. Return: 36.3% (SD = 21.5%)
    CAGR: 35.1%

     

    With UBT:
    Win Rate: 69.2%
    Median Rolling 12-Mo. Return: 37.6% (SD = 36.3%)
    CAGR: 49.0%

     

    With TMF:
    Win Rate: 65.4%
    Median Rolling 12-Mo. Return: 21.1% (SD = 53.1%)
    CAGR: 44.2%

     

    So at face value, adding leverage to the long-term treasury ETF might improve results a bit, but also introduces greater volatility. I didn't calculate Sharpe ratios, but my guess would be that substituting UBT and TMF would diminish Sharpe ratios because of the greater volatility.
    11 Aug 2014, 07:18 PM Reply Like
  • crakes
    , contributor
    Comments (4) | Send Message
     
    The key to rotational strategy success is the etf universe. Dave Varadi has blogged on one method to identify what is needed to have a well constructed universe for rotation. How did you go about selecting your universe(s) for this return - trail and error or is there a systematic process?

     

    Thanks!
    2 Aug 2014, 10:53 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Sorry, I should have mentioned that the ETF basket was chosen by Mr. Grossman in an earlier very popular article on SA.

     

    http://bit.ly/1mEMAoa

     

    I just assumed that this will be obvious to the readers.

     

    I just wanted to provide a simpler version that does not use volatility for rankings, as it seems to me that the optimization that might have gone into choosing just the right combination of relative strength and volatility may not necessarily be the best way to develop a robust strategy.
    2 Aug 2014, 02:36 PM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    Hello Varan,

     

    It is pretty clear that you were basing your ETF choices on Frank Grossman's article.

     

    Now that I think about Grossman, I was wondering what this strategy would look like under your thoughtful lense: http://seekingalpha.co....
    2 Aug 2014, 02:55 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @algyros Most of those strategies are based on EDV for which the date of inception is quite recent (2009). For the 2009-todate, MDY/EDV :21.7%, ZIV/EDV 40%.
    2 Aug 2014, 07:03 PM Reply Like
  • rorubs
    , contributor
    Comment (1) | Send Message
     
    Thanks for sharing your twist on grossman/cohn strategies.

     

    picking, I would be careful with the - always possible, especially with these strategies - hindsight bias in selecting the universe.

     

    Especially, ijj seems a little odd compared to the other etfs (mid caps vs big caps, which is maybe better played with another strategy), ilf overweights Latin America vs Eastern Europe or Emerging Asia, and Japan is excluded. In hindsight, those choices make sense, but without it... For robustness purpose, I would suggest to test adding spy/Japan and removing ilf/iff

     

    That being said, it seems treasuries are not rotated into very frequently, which might be good for the future.

     

    Cheers.
    2 Aug 2014, 11:01 AM Reply Like
  • IndyDoc1
    , contributor
    Comments (312) | Send Message
     
    How about shorting the corresponding inverse 3x Direxion fund of the winner to capture the beta slippage. For example, the winner this month of Grossmann's GMR is EEM, how about shorting EDZ!
    2 Aug 2014, 06:46 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » That - shorting EDZ when EEM is the ETF to be invested in - may well work. But I have not tried it.
    2 Aug 2014, 07:07 PM Reply Like
  • EdwardjK
    , contributor
    Comments (197) | Send Message
     
    varan,

     

    I assume the three month measurement period is 63 trading days, yes?

     

    Can you devise some circuit breaker that moves the monthly investment to cash in anticipation of a market decline? maybe something as simple as an x% decline within x trading days…

     

    Thank you.
    2 Aug 2014, 08:05 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » It's three calendar months. So for January: last trading day of prior year's September thru last trading day of the prior December, and then increment both the months by a month for subsequent months. My guess would be that using 63 days should work as well. If there is a substantial difference, I will withdraw this whole strategy as that would imply the strategy to be very brittle.

     

    The circuit breaker is a good idea, but I have not implemented that yet, and therefore not tested it.
    2 Aug 2014, 08:28 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    A circuit breaker or trend filter is an interesting idea. For GMR "simple" I tested the following filters:
    - Mom > 0;
    - Close > SMA4;
    - Close > EMA4;
    - SMA4 > SMA4[-1];
    - EMA4 > EMA4[-1];

     

    The length of the simple and exponential moving averages is based on the lookback period varan used in his test: 3 months, hence 4 bars. All versions of the trend filter lead to lower CAR, Sharpe and Sortino, except for the SMA4 > SMA4[-1] version. Applied to monthly quotes this filter compares the current (or last) value of a 4 bar/month simple moving average against the next to last reading, resulting in a slightly higher CAR, but more importantly: a 2.5% better max draw down percentage.

     

    The next link shows the comparison overview:
    http://bit.ly/1sNrLsr

     

    The equity curve with key perfomance indicators for GMR simple with the SMA4 > SMA4[-1] trend filter:
    http://bit.ly/1sNrK7R

     

    This exercise with trend filtered momentum rotation brings an older comment into mind made by a visitor on my blog:
    "(...) I am not very enthusiastic about the way (...) momentum [is measured]."
    and
    "You can pick any TF function, from something as simple as Mebane's to the more sophisticated one Andreas Clenow describes in his TF book (...) and TF will figure it out automatically what works (and what doesn't)."
    (For more, see: http://bit.ly/1sNrLsv)

     

    If anyone knows a better way to measure momentum or apply a trend filter, please step forward.
    12 Aug 2014, 11:09 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    I use SHY and 200 day moving average as circuit breakers. If no funds are above 200 day average or SHY is the best performer for the last three months then monies go to cash.
    2 Aug 2014, 09:07 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I have not tried the SMA, but adding SHY degrades the performance a bit: CAGR decreases by about 2%, and the maximum drawdown increases by about 3%, and the Sharpe decreases by 0.08. But even then it may be a good addition as a circuit breaker.
    2 Aug 2014, 09:21 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    I just finished a twenty year backtest using FUSEX and VUSTX. Updating monthly with the one that has the best relative strength for the last three months. performance was almost double and declines were reduced. I am currently going to rework that study to go to cash when both are negative to see how this will work. You might be able to do it very quickly with your math skills.
    2 Aug 2014, 11:10 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » That is paired-switching. For 1991-2014 you get a CAGR of 12.7% for FUSEX( Fidelity does have a sense of humor)/VUSTX. The best pair is FDVLX/VUSTX: if you update quarterly based on just prior month's returns you get 16.4% for 1991-2014. The GMR, in my nomenclature, is polygamously paired switching.

     

    .
    3 Aug 2014, 12:14 AM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    Why would you limit yourself to a pre-defined shortlist of ETFs or mutual funds? Since it's not a buy & hold or volatility based strategy why not trying to max your gain as long as the momentum is there?

     

    You may want to create ETF groups only including well-traded ETFs and make sure there's not for instance 5 Indian ETFs in one group, but for the rest I'd say buy ANY ETF that has been gaining above the cut off point (2%?, SHY?), is above the 200 DMA and is gaining momentum. And you could set a selling rule like going to cash for the rest of the month if there's a 3 or 5% loss compared to the buying price.

     

    In this way you don't have to leave out levered and/or inverse ETFs and you don't care about asset allocation, economic climates, market conditions, crashes, et cetera. You just buy the ETFs that have momentum.

     

    Note: For levered and/or volatility ETFs you may want to cap your gain, like take a 5% monthly gain and get out, or use a trailing stop.

     

    I'm relatively new to this subject but after reading 50 or so articles about this subject it isn't clear to me why you would restrict yourself if you're focusing on momentum without theoretical "life savers" like risk parity, relative strength, et cetera. After all, who cares about this if you have an exit strategy that limits losses?

     

    Thanks,

     

    drftr
    6 Aug 2014, 04:10 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Of course. There are multiple ways to invest successfully. A method like this provides one of many systematic approaches that is easy to implement.
    6 Aug 2014, 09:53 AM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    If you backtest a strategy such as Varan's (for example, on ETFscreen.com), you'll find that if you use too many ETFs the results suffer. In fact, if you use the universe of all ETFs on their database, the results are awful.
    6 Aug 2014, 10:33 AM Reply Like
  • vn888
    , contributor
    Comments (12) | Send Message
     
    @Algyros:

     

    I think David Varadi has a good explanation why this is the case in his blog post on Momentum Score:

     

    http://bit.ly/X2AFZi
    6 Aug 2014, 06:13 PM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    I read a great line from a book on fly fishing by Yvon Chouinard that said that performance genius/mastery is in finding not in what one can add, but in what one can take away. I think Varan implicitly understands this. That is the genius of his ways, take it or leave it.
    6 Aug 2014, 10:20 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @ikkyu
    thanks a lot.
    6 Aug 2014, 03:27 PM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    I hear what you say Ikkyu and the portfolio that is still most appealing to me is something like trading the best 1 or 2 out of VTI-TLT-GLD-SHY as this will do very well in about any market situation. But for an other part of my investments I'd like to find out more about the limits of momentum trading. And therefor all responses are educative to me. Tnx!

     

    drftr
    6 Aug 2014, 10:59 PM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    Varan,

     

    Do you have any selling rules in place or are you always waiting until it's time to decide which ETF to invest in for the next period of time?

     

    drftr
    7 Aug 2014, 09:54 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @drftr

     

    I do not have any selling rules.

     

    If there is a change in the selection on the first trading day of the month, the one held in the prior month is sold, and the proceeds are used to invest in the new one. Otherwise, no action is necessary.

     

    Thanks.
    8 Aug 2014, 12:02 AM Reply Like
  • IndyDoc1
    , contributor
    Comments (312) | Send Message
     
    How about placing your favorite FLPSX instead of FDVLX ?
    8 Aug 2014, 12:45 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » that will work too, probably better. I think though that FDVLX behaves more like other value funds, and FLPSX is more active.
    8 Aug 2014, 12:51 PM Reply Like
  • TrickPony
    , contributor
    Comments (34) | Send Message
     
    Is anyone using this or similar momentum type systems with real money? if so what percentage of your portfolio are you comfortable allocating too it?
    9 Aug 2014, 01:48 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    I track 40 ETFs and try to stay invested in the top 5-10 and out of or short the bottom 5-10. 55% of my portfolio is involved with this strategy at the present time.

     

    I use 4 month relative strength for most of my calculations.
    10 Aug 2014, 10:03 AM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    extremebanker,

     

    How did you select those 40 ETFs? Personally I would be looking at the top ETFs for any economic climate (Varan may for instance want to change his ETFs if/once high inflation kicks in) so you can ride the trend from an early stage, but I'm interested to find out what moves other investors. Tnx!

     

    drftr
    11 Aug 2014, 09:17 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I haven't seen or done an analysis, but if you use relative strength with a basket of very large number of funds, the probability of false discovery during backtesting might increase. This is speculation on my part, of course. Even if it is correct, no one knows how large large is in that context. Clearly, since it is working well for extermebroker, the threshold is perhaps much higher than 40.
    11 Aug 2014, 09:34 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    Nine of the ETFs are large, mid, small growth,blend and value. Two are domestic and international real estate. Two are commodity and gold related. Five are international stock related. Seven are related to fixed income. Ten are actually funds that are in my 401k. The remainder are country specific and sector specific chosen from a list. A good source for relative strength strategies is "quantpedia"

     

    When my performance really excels is when the bear markets strike. I go to cash and treasuries which is a good mix in a bear market.
    11 Aug 2014, 06:05 PM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    Excellent - thanks! And do you limit your choice by only selecting ETFs with a minimum turnover, minimum years on the market, maybe even buying and holding prices?

     

    I'm trying to figure out a sensible way to get rid of many 100s of ETFs that would fit the bill but may be to risky.

     

    drftr
    12 Aug 2014, 07:42 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    I prefer ETFs with at least 500 million in assets. I also want funds that represent distinct asset classes. I try not to have too much duplication. A simple version is included with the link below.

     

    http://bit.ly/1sNeiAZ
    12 Aug 2014, 10:20 AM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    Either the link is not working or it's blocked by the Chinese government (which may be the case if there's anything Google-ish in there). I'll find a way around it in a couple of days in another city and have a look. Thanks for posting.

     

    drftr
    12 Aug 2014, 10:35 AM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    The link works for me (I am in the US).
    12 Aug 2014, 04:07 PM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    A VPN connection in China makes everything on the web work again... Thanks extremebanker for the link.

     

    drftr
    13 Aug 2014, 09:32 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » The August selection EEM has returned 2.27% todate. The final return will be determined by the closing price on 9/3/2014.

     

    For September, ILF is the clear selection, as its return during the period 5/30/2014 thru 8/30/2014 has been 15%. The return of the second ranked ETF during this period for the un-leveraged basked was 6.7% for EEM and for the leveraged basket it was 8.9% for MVV.

     

    Summary:

     

    8/2014 return: 2.27% (both baskets)
    9/2014 selection: ILF (both baskets)
    9/2014 second ranked ETF:
    EEM (unleveraged basket) return 6.7%
    MVV (leveraged basket) return 8.9%

     

    (All data based on Yahoo Finance adjusted close prices.)
    31 Aug 2014, 02:28 PM Reply Like
  • Steadyprofit
    , contributor
    Comments (42) | Send Message
     
    Thanks for this months update. Do you have any hesitation or timing suggestion about rolling over into the new selection of ILF after it had such a stellar return this past month? EEM did well this past month but Mr. Putin's adventurism seemed to put a damper on it the last few days; even so I was happy with my results. Just struggle each month as to the "best" method of transitioning into the next selection.
    1 Sep 2014, 06:05 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Thanks.

     

    I will leave that subjective judgment to anyone who wants to actually implement it.

     

    I would just like to follow this for a while, unadorned by any subjective criteria, to see how it works out.

     

    Given that reservation, I am going to buy ILF tomorrow, solely on the basis of this ranking.

     

    Sorry.
    1 Sep 2014, 06:13 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Korndog, just my opinion, but I think when you begin to second guess things, that defeats the purpose of an algorithmic strategy such as this. That's a slippery slope that invites introducing more and more subjective bias.

     

    It may be tempting, but more often than not "the enemy of good is better."

     

    Even though ILF has had a runup and is testing moderate chart resistance, there are no signs of technical weakness that I can see. The ETF is overbought to be sure, but overbought conditions are not necessarily unstable and can persist indefinitely.
    2 Sep 2014, 01:28 AM Reply Like
  • TrickPony
    , contributor
    Comments (34) | Send Message
     
    Wow...ILF took a pounding today.
    Kurt
    8 Sep 2014, 04:12 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Yes it did. Now we can see significant technical damage and sell signals in place. But that said, TAA is not based on technical indicators, so we'll just have to see how this plays out over the rest of the month.
    8 Sep 2014, 09:02 PM Reply Like
  • TrickPony
    , contributor
    Comments (34) | Send Message
     
    I think we are approaching the 200 day moving average, if we break that look out below.
    23 Sep 2014, 01:40 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yeah. Getting ugly.

     

    Given that during the period Jan 2003 thru Aug 2014, only 2 months out of 155 had losses of more than 10% and only 7 losses greater than 7.5%, this seems like a very uncharacteristic month for the strategy.
    23 Sep 2014, 02:05 PM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    Very true! I run with 10% stop loss circuit breaker on my momentum positions, because, as you say, such triggers are rare and have meaning.

     

    The market might be rolling over. I lot of movement all of the sudden.

     

    Cheers.
    23 Sep 2014, 08:41 PM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    EDV has pulled to the front on the maximum yield rotation strategy. Will see if he holds through to the switch at the end of the month.
    23 Sep 2014, 08:45 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Hi TP,

     

    ILF blew through the 50sma and 100 sma like a hot knife through butter. But now it is extremely oversold, has gotten away from its 10sma, and looks ripe for a mean reversion sort of a bounce here.

     

    If so, then the next big question is whether ILF can get back up to the prior peak that it hit about a week ago at $41. That's the critical test in my view, because it ILF cannot make it up that far, then that means it will have formed two lower lows and two lower highs. That would be a very bad sign indeed, and a test of the chart support region around the 200sma would be the next step.

     

    Personally, I thnk the 200sma carries little significance in the case of ILF. If you look at the long-term chart, I think you will see that ever since the bubble implosion of 2008, ILF has been stuck in a wide trading range, and has repeatedly violated the 200sma with impunity, but that never seems to indicate an orderly rolling over of the trend. Conversely, when ILF rallies through the 200sma, that doesn't necessarily mean that everything's wonderful and ILF will just steadily churn upwards. That's just not the nature of post-bubble ILF anymore, which is invariably the case with markets that are unwinding bubbles.

     

    For reference purposes, just look at China, which is pretty classic post-bubble behavior. It is now about 7 years past the peak, and the Shanghai and Shenzhen composites are trading 65% below their peaks.

     

    That market is particularly striking from a chartist/technician's standpoint, because while the underlying economy has been expanding strongly, equities have just disintegrated, show no signs of life now, and in my opinion haven't hit rock bottom and won't equal the previous highs in real value terms for another 20 years at least.

     

    Anyhow, bottom line is that I see an imminent bounce to be very likely, but no buy signals yet.
    24 Sep 2014, 01:25 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    Paired switching and some other rotation strategies (IVY portfolio) have underperformed the last several years. I believe they have a tendency to underperform in bull markets but make up for it when the bears strike.
    23 Sep 2014, 02:36 PM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    It is EDV for Global Market Rotation and Max. Yield Rotation for Oct. based on performance only.

     

    Curiously, ETF tools showed SHY for GMR with the 70/30 (return/vola) calculation.

     

    I think Varan is a TLT, rather than a EDV, man.

     

    Cheers from Osaka,

     

    john
    1 Oct 2014, 01:00 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yes, TLT it is.

     

    Return Aug 1/2014-todate: -12.2% Yuck.
    1 Oct 2014, 01:12 AM Reply Like
  • jeezuz30
    , contributor
    Comments (607) | Send Message
     
    I wonder if the strengthening dollar had much to do with ILF's performance this last month.

     

    At least the EWZ has been getting killed and that's probably the biggest component of ILF. Maybe next time one of these volatile ETFS like ILF or EPP come up as first place it may be better to split it with either EDV or MDY to control the high volatility.
    1 Oct 2014, 12:20 PM Reply Like
  • Steadyprofit
    , contributor
    Comments (42) | Send Message
     
    I think you are right on both accounts. The volatility cuts both ways, and sure wasn't much fun this past month. I'm pretty cash heavy right now, so it wasn't horribly painful, but I still don't like loosing. ILF can be a wild ride and may need to be tamed a little even at the expense of some upside.
    1 Oct 2014, 12:59 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Korndog and jeezuz, I am not sure normalizing for volatility will help.

     

    There was discussion here about this maybe 5 or 6 months ago. As I recall, both Varan and Cohn tried normalizing for volatility, and it didn't help returns and probably hurt.

     

    With 20/20 hindsight, it would have helped Sept results, but as you say, that's a double-edged sword, and it will also tend to blunt upside returns.

     

    Terry
    1 Oct 2014, 04:05 PM Reply Like
  • SpanglerDavis
    , contributor
    Comments (1008) | Send Message
     
    In my experience, the building blocks are:

     

    Find a switching or cutoff mechanism for downside risk
    Look at pair switching for return when momentum warrants
    Select a frequency of calculation that is appropriate for the universe of ETFs chosen
    Only add those ETFs to the mix that provide meaningful and consistent volatility.
    Too big of a universe results in lower returns
    Erratic ETFs cause lower returns through no follow through on momentum.
    For example, IBB is a tough ETF to include due to its extreme violatility and the eventual mistakes made by the system.
    6 Oct 2014, 10:42 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » good summary. thanks.
    6 Oct 2014, 10:57 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Positive return for 10/2014 but lower than SPY.

     

    The selection for November continues to be TLT.

     

    I will follow this till the end of 2015 at least.
    2 Nov 2014, 11:43 AM Reply Like
  • TrickPony
    , contributor
    Comments (34) | Send Message
     
    Thanks for the update
    3 Nov 2014, 02:56 PM Reply Like
  • Activelypassive
    , contributor
    Comments (7) | Send Message
     
    Whats the update for Nov 14? still TLT?
    1 Dec 2014, 11:00 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yes, still TLT.
    1 Dec 2014, 11:30 AM Reply Like
  • Activelypassive
    , contributor
    Comments (7) | Send Message
     
    Cheers Varan.
    1 Dec 2014, 12:42 PM Reply Like
  • alsobirdman
    , contributor
    Comments (433) | Send Message
     
    Hi Varan,

     

    Just found you while reading Frank Grossman's latest. I want to be sure I understand this correctly;

     

    For the RS-GMR-ETF: IJJ, IEV, ILF, EPP, EEM, TLT - ETF strategy, you are 100% in one of the 6 funds, and that one fund is the one that was the best performer over the previous 3 months? It seems so easy I figure I might be over-thinking it and miss something?

     

    Thank you
    2 Dec 2014, 04:24 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » That's correct. On 1st of January (and so on for the first trading day of every month), I find the ETF whose total return (based on the adjusted close data from Yahoo Finance) during the period Oct thru December of the prior year.

     

    Starting in January 2015, I will add QQQ and drop ILF, as this improves the year to year volatility of the returns.
    2 Dec 2014, 04:28 PM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    Which would mean QQQ for December if you did the change yesterday, correct?
    2 Dec 2014, 04:42 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yes. That is correct. Thank you for your interest. I will post a detailed note in the next few weeks.
    2 Dec 2014, 04:49 PM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    thanks, v. I'll be looking for it.

     

    I've found this exercise interesting. From following several momentum stories late last year I decided to move a bunch of my fixed-income allocation to TLT in the spring for a month, and have had it there ever since. Without having been looking at momentum articles I would never have considered TLT.

     

    Not sure how it stands today, but as of a few days ago TLT had been beating SPY and QQQ soundly YTD.

     

    It got my attention and I've been studying up on momentum strategies.

     

    Final point: I notice a lot of momentum investors do incorporate a volatility component. Haven't seen much use of that here and wonder if you have any thoughts.

     

    Thinking back a few months when this portfolio went to ILF (I've been watching but not actively investing) might that have prevented that tough month?
    2 Dec 2014, 05:09 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I have incorporated volatility only in the risk parity based balanced strategy that I have described in a prior post.

     

    Perhaps some combination of volatility and return would have excluded ILF from the GMR selection. I don't know if that is worth finding out.

     

    Thanks.
    2 Dec 2014, 08:11 PM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    I have never been able to backtest it, but choosing the momentum fund pick basked on the fund with the highest Sharpe Ratio seems like simple way to include volatility risk.
    3 Dec 2014, 03:36 AM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    To touch base quickly: the charts for GMR with QQQ instead of ILF.
    http://bit.ly/1AvUAxO
    5 Dec 2014, 02:53 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Thank you TrendXplorer.

     

    Nice charts.

     

    With ILF replaced by QQQ:

     

    CAGR 27.15% maxDD 14.6% Sharpe 1.38 Sortino 2.94

     

    Without Replacement

     

    CAGR 26.4% maxDD 17.1% Sharpe 1.21 Sortino 2.41

     

    More importantly, and the results that dictated the modifications, the recent annual returns (the modified GMR results in parentheses)

     

    2012 6.5% (15.6%)
    2013 20.7% (24.2%)
    2014 0% (25.0%)
    5 Dec 2014, 03:05 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @Trend:

     

    Very nice JW, thanks!

     

    I notice that rolling returns plunged at the worst possible time (mid- to late-2008), but then again the DD was nothing compared to what most people were experiencing at that time.
    5 Dec 2014, 10:34 PM Reply Like
  • Market Map
    , contributor
    Comments (843) | Send Message
     
    Don't you think that replacing ILF with QQQ puts too much tilt ( combined with IJJ ) in the whole thing towards a U.S. weighting ? There certainly are theories about why Latin America is having problems but, even as the ILF has had a string of deep and discouraging losses and "momentum" failures, sometimes these problems resolve themselves in surprising ways and the ILF could roar back with a string of significant gains. This does seem to happen in these type of systems; a big loss comes through, we get disconcerted and try to rework with the formula, and then miss out on the gains from what the old version would have given us if we were able to "stomach" our disconcertion.

     

    Viewing the breakdown of GMR aggregate historical monthly returns , ILF is still the leader since 2003 here: http://bit.ly/1vYcitL
    5 Dec 2014, 03:45 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I am with you on your basic thesis and the methodology.

     

    I included QQQ mainly for pragmatic reasons - three years of less than mediocre performance is quite a long time.

     

    I will personally use the modified version but continue to track both.

     

    Thanks.
    5 Dec 2014, 04:19 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @Market Map:

     

    Hmmm...I would say yes and no.

     

    On one hand, I don't think it matter much how much weight is in what market. Varan (and others) have TAA strategies that begin with baskets of securities that represent sectors within the S&P500, and these can work quite well.

     

    Instead, I think one of the key factors is that the securities in the starting basket are variably correlated/uncorrelated with one another. The ideal situation is where the correlation cyclically varies over a range from -1.00 to +1.00. Look at the 20-day correlations of the QQQ:ILF or IJJ:ILF ratios and compare those to the 20-day correlations of the QQQ:IJJ ratios:

     

    http://bit.ly/1q4ol87

     

    http://bit.ly/1q4olok

     

    http://bit.ly/1q4okkk

     

    I think you will agree that the former two pairs generally tend to vary over a wider range and with a more regular periodicity (or relatively constant cycle length) than the QQQ:IJJ ratio.

     

    This tells me that adding QQQ and ILF to the basket would create a better mix of ETPs than adding QQQ and IJJ, but not for the reasons you suggest.

     

    Why should correlation matter? Because that enhances the probabilities that a momentum-based strategy with look-back periods that are of a reasonable length compared to the cycle length of the varying correlations will detect when relative momentum shifts from one ETP to the other. It doesn't help to have a good strategy to detect when the momentum baton gets passed from one ETF to another if that baton never gets passed, and the ETFs just move in lockstep most of the time.

     

    I am also suggesting that the magnitude of the range of correlations is important as well: the wider the swings, the better. That way, the signal:noise ratio is enhanced, and your strategy has a better chance to detect signal and not be unduly swayed by noise.

     

    Fortunately, the whys and wherefores and geopolitical and macroeconomic inputs that might be causing the shifting correlations and relative momentums are all but irrelevant. The centrally important facts are that they happen, can be detected with reasonable sensitivity/specificity, and can therefore potentially be exploited for profit.

     

    TMD
    5 Dec 2014, 10:56 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    Starting ultimo 2002 with a hypothetical $100,000 as initial capital and with IJJ, IEV, ILF, EPP, EEM and TLT as our portfolio, at the end of last August ILF's profit contribution still was at a decent 12%:
    http://bit.ly/15XTlOf

     

    September's severe hit for ILF changed its total contribution to -2%:
    http://bit.ly/15XTlOh
    So in hindsight the portfolio would have done better without ILF at all.

     

    These percentages are taken from each ETF's profit contribution measured in $-dollars compared to the portfolio's total dollar profit (End November: $2.095.000).

     

    While to a certain extent the severity of September's blow to the portfolio remains concealed in ILF's aggregate monthly returns, your trading account would have uncloaked it for sure: a $295,000 setback.
    http://bit.ly/15XTj8X
    (check the last column sorted on loss in $-value)
    7 Dec 2014, 04:33 PM Reply Like
  • SpanglerDavis
    , contributor
    Comments (1008) | Send Message
     
    Why replace ILF with QQQ? Just add QQQ to the universe from which you choose. The most important attribute is that you only select one investment per time period. The issue then becomes do you want 100% of your assets in Latin America or the 100 largest US companies? If you want EEM or ILF in the universe then the holdings should increase to either 2 with some form of volatility weighting or to a 2 ETF portfolio where a certain percentage is allocated to US equities with EEM or ILF as a satellite holding when appropriate.
    5 Dec 2014, 06:16 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Good idea. I will take a look.

     

    Thanks.
    5 Dec 2014, 06:21 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » During 12/2014, TLT (the original Simple GMR selection) returned 3.8%, whereas QQQ (the selection for the same basket but with ILF replaced by QQQ) lost 1.1%.

     

    These would have been the yearly returns for 2014:

     

    Simple GMR: 3.6%
    Simple GMR with QQQ: 23%.

     

    2003-2014 Performance

     

    Simple GMR: 26.8% Worst Year: 2014 (3.6%)
    Sharpe 1.24 MaxDD 17.1%

     

    Simple GMR with QQQ: 27% Worst Year: 2004 (13.6%)
    Sharpe 1.38 MaxDD 14.6%

     

    For 1/2015, the selection for both is TLT (aargh!).
    2 Jan 2015, 01:39 PM Reply Like
  • Stanford Chemist
    , contributor
    Comments (1974) | Send Message
     
    Thanks for the continual updates Varan. Interesting that QQQ gives such a big improvement over ILF.
    2 Jan 2015, 01:49 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    I have found broad indexes seem to work better than using sector indexes or region or country funds. In other words, EFA and EEM work better than using EPP,ILF and IEV.

     

    Much more volatility with funds that are too specific.
    2 Jan 2015, 01:56 PM Reply Like
  • Activelypassive
    , contributor
    Comments (7) | Send Message
     
    Thanks Varan for all your updates..
    5 Jan 2015, 07:57 AM Reply Like
  • speranz1
    , contributor
    Comments (17) | Send Message
     
    Thank you to Varan for following this strategy as well as sharing several innovative allocation strategies. Most, if not all, have been well thought out and insightful.

     

    One question I'd like to raise: when evaluating which funds to use in our "basket" or monthly allocation, to what degree do we test whether we are backtest overfitting? I know this is a topic that Varan and others have touched on in the past, but increasingly worries me when implementing this strategy.

     

    To me, and this is purely subjective, the substitution of QQQ for ILF seems to violate the investment thesis, or at least raises a red flag that I'd attempt to resolve before investing. The reason mid-cap US equities was used in the original GMR was to provide an allocation to US equity that had a relative volatility in line with the global equity funds. Using QQQ, at least on the surface, seems to violate that reasoning.

     

    Additionally, in my mind this strategy works to some degree due to global supply and demand forces. Momentum in and of itself is not an investment hypothesis, but relative momentum within a basket of global equities (plus the often forgotten TLT) due to long and short term swings in demand for equities in a given global market is. Removing ILF for QQQ would violate or at least challenge such a thought process by narrowing the global exposure.

     

    I guess what my question boils down to, is how do we evaluate whether the substitution of an asset is merely overfitting the backtest? And depending on how we resolve that question, how do we evaluate whether Grossman's strategy does the same?
    12 Jan 2015, 04:29 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I share your concern in general.

     

    Although the modification was prompted by mundane pragmatic considerations, it is not entirely without analytical basis.

     

    Inasmuch as EEM does contain some exposure to Latin America, the ILF could be a bit redundant in any case.

     

    This is borne out by the fact that if you look at the correlation matrix of the basket, the largest (non-diagonal) element is the cross-correlation between EEM and ILF.

     

    In addition the smallest off diagonal elements - when you include QQQ - in the correlation matrix are the cross-correlations between QQQ and EPP and EEM.

     

    I did not do any search for the best replacement, and just tried QQQ, and, serendipitously, it proved to be quite good.

     

    I know that this is not as satisfactory an answer as the issue demands, but probably suffices as the first step.
    12 Jan 2015, 05:35 PM Reply Like
  • jeezuz30
    , contributor
    Comments (607) | Send Message
     
    Are you still using 3 months for your look back period? Is this still the optimal timeframe or have you made any adjustments?

     

    Thanks
    18 Jan 2015, 05:00 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Just three months. It may well be that some weird number of days like, say, 53 or 41, will provide a better return for the past data, but I take all such optimized strategies with a huge grain of salt as they may not be very robust. Most of the time I only test for three and twelve months of lookback for annual strategies, and only three months for the ones with shorter holding periods.

     

    Thanks.
    18 Jan 2015, 05:11 PM Reply Like
  • jeezuz30
    , contributor
    Comments (607) | Send Message
     
    Thanks for the reply, good to see that you can remain with a strategy for quite some time without the need to constantly optimize.
    19 Jan 2015, 02:56 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » As much as I would like to get out of TLT, its returns continue to dominate those of other assets. It has, however, saved me from some minor losses during January (TLT returned 8.6%, and January returns of others range from -2.7% to 1%).

     

    The last three months' returns for the top three ETfs in both baskets:

     

    TLT 16%
    QQQ .1%
    IJJ -.4%

     

    So if GMR is to be followed mechanically, no change is required for the month of February for either of the strategies. Common sense and simple arithmetic suggest that one of these months TLT will underperform and lead to a loss, but when I do not know.
    1 Feb 2015, 01:04 PM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    Not to be too flip, but a 7.6 to 11.3 percentage point differential in a month's time hardly fits my definition of minor. Even more so 15+ percentage point differential over the next best two for 3 months.

     

    Agree that it seems inevitable that TLT will drop, but I've been telling myself that for months now. Right now, I don't share your reluctance to stay that course. Isn't that what a momentum strategy dictates? Anyway, it's going to have to crash a lot -- and a lot more than I think it's slated to given the macro factors that are driving its rise -- to make me regret my willingness to keep on keepin' on.

     

    At worst I expect it to flatten. The rest of the world will continue to buy LT treasuries for one good reason: there's no place else to go. Look at what happened after it dropped on Thursday. Came roaring back with an opening gap that not only sucked up that entire loss but set a new high yet again.

     

    Of course all of this might well mean it's time to bail. After all, one of my commenters compared my views to getting tips from your shoeshine boy.
    1 Feb 2015, 07:47 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Thanks Leftbanker.

     

    I think I did not communicate well. I agree with your point. I was just trying to make sure that anyone who follows this understands the uncertainties, with which I am myself completely at ease.
    1 Feb 2015, 08:11 PM Reply Like
  • Activelypassive
    , contributor
    Comments (7) | Send Message
     
    Hi Varan

     

    Is the market movement in March changed the rankings?

     

    look forward to you updates.

     

    thanks
    2 Mar 2015, 07:37 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Return of TLT for the period 11/30/2014 through 2/27/2015 was highest among all the ETFs in the original basket. It was higher than that of QQQ as well.

     

    So for both of the methods TLT is still the selection for 3/2015.

     

    Bummer.
    2 Mar 2015, 10:02 AM Reply Like
  • Activelypassive
    , contributor
    Comments (7) | Send Message
     
    Varan

     

    There could be a change from TLT to IEV, do you think this is the case?

     

    Cheers
    1 Apr 2015, 06:02 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yes. That's correct.

     

    The prior three months' return of both is almost the same (4.2% TLT, 4.1% IEV), with EPP's return being 3.2%. However, if you compute the return for 2015/1/2 thru today, IEV is higher. I will go with IEV.

     

    YTD return 2.8%.

     

    Thanks.
    1 Apr 2015, 09:29 AM Reply Like
  • Activelypassive
    , contributor
    Comments (7) | Send Message
     
    you are a star as ever!
    1 Apr 2015, 10:31 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » IEV has done ok for April.

     

    For May, EEM is the selection.

     

    YTD: TLT 1/2/2015-4/1/2015 4.4%
    IEV 4/1/2015-todate 3.18%

     

    YTD Return 7.42%

     

    For comparison: VTI 2.28% EEM 9.1% IEV 8.35%
    30 Apr 2015, 10:49 PM Reply Like
  • nimman919
    , contributor
    Comment (1) | Send Message
     
    Hi Why it is not include a japan markert?
    4 May 2015, 03:43 AM Reply Like
  • user208606
    , contributor
    Comments (3) | Send Message
     
    In case I missed this, is the ETF purchased based on the closing price on the first day of the month?
    10 May 2015, 12:37 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » yes.

     

    the selection is based on the returns for the one month period ending on the last trading day of the previous month, and the transaction is made at close on the first trading day of the current month.
    10 May 2015, 12:50 AM Reply Like
  • user208606
    , contributor
    Comments (3) | Send Message
     
    Thanks for the quick response. A follow up question; If EEM is used for the month of May and the end of the month dictates a change to another security (based on prior 3 months performance of the ETFS ( IJJ, IEV, QQQ, EPP, EEM, TLT), would the sale of the ETF also occur at the same time when the purchase is made? ie. MOC Sell OLD, MOC Buy NEW on June 1st. I have seen models on etfreplay where the sale and purchase dates seem to be make a dent in overall returns. Thanks again for sharing your strategies.
    10 May 2015, 02:17 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yes. Though at least my broker does not allow transactions in share counts other than in multiples of 100, all the results here are based on both the transactions occurring at market close on the first trading day of the month so those who wish may verify the results.
    11 May 2015, 02:50 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    According the GMR Simple App (under construction) using varan's original six funds, the pick for June is IEV:

     

    http://tinyurl.com/ohp...

     

    The screenshot also shows that May's pick EEM yielded a loss of -4.1%.
    31 May 2015, 08:03 AM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    TrendXplorer,

     

    How about positioning a part of your great blog for updating all SA momentum strategies? If something like that could be updated automatically you'd do the community a great service AND attract some more readers. I would happily pay for that service so a Donate button would serve you well.

     

    UNLESS the portfolio designers don't agree of course, but I can imagine it's a pita for some or them.

     

    drfrr
    31 May 2015, 08:11 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » From close of market 5/1 thru today I see a loss of 4.68% from Yahoo finance.

     

    YTD 3.5% (TLT 1/1 trhu 4/1, IEV 4/1 thru 5/1, EEM 5/1/)
    SPY 3.18%
    DIA 2.08%
    EFA 9.5%
    31 May 2015, 10:41 AM Reply Like
  • C.J.102929
    , contributor
    Comments (59) | Send Message
     
    Hi Varan,
    Is it EEM or IEV? Thanks!--c.j.
    15 Jun 2015, 01:21 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » It was IEV on 6/1 as Trendexplorer stated. Thanks.
    15 Jun 2015, 01:39 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » By the way my strategies are truly 'open source'. Use them in any way you please with your own due diligence and at your risk. No need for giving me credit either as it will not do me any good. As I say I just want to share the results that I think are interesting and presume that a few others may find them interesting too
    15 Jun 2015, 01:43 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    By popular demand I published a page with "live" signals for the Simple GMR strategy:
    http://bit.ly/1IgkYO9
    Use the signals in the spirit of varan's above remarks.
    Please take note of the slightly different approach: reforms are at the close of each month.
    (A post about the real risk involved in these kind of strategies is pending, but most likely won't be published before the summer ahead)
    15 Jun 2015, 04:17 PM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    I agree with drftr. That would be awesome.
    31 May 2015, 10:28 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Pretty bad month. IEV down 2.8%, so the YTD return is essentially zero (.2%).

     

    For July, the selection is ILF for the original basket, and QQQ for the modified basket. For July onwards, I am adding EFA to the modified basket, which now contains IJJ, QQQ., EEM, EPP, IEV, EFA, and TLT.
    1 Jul 2015, 01:51 AM Reply Like
  • CostBasis
    , contributor
    Comments (12) | Send Message
     
    Have you considered adding IWC (micro-caps)? My thinking is that if you believe in high-beta, it is like IWM (Russell 2000) on steroids. If IWC was in the mix, it would be selected for July over ILF and QQQ.
    1 Jul 2015, 03:08 AM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    Update: the "live" signals are now in accordance with the new composition of the GMR Simple universe:
    http://bit.ly/1IgkYO9

     

    To add some perspective for the 7 asset basket:
    - profit table: http://bit.ly/1f0zhQ5
    - allocations: http://bit.ly/1f0ziTX
    Results are with total return quotes provided by PremiumData.

     

    @CostBasis: with IWC in the basket too, both CAR and MaxDD drop by about 3%.
    1 Jul 2015, 08:03 AM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @TrendXplorer,

     

    Great work, as per usual. Thanks very much for this!

     

    TMD
    1 Jul 2015, 07:28 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » QQQ did reasonably well last month, ILF not so much.

     

    From now on, I will just be tracking the GMR for the following basket:IJJ, QQQ., EEM, EPP, IEV, EFA, and TLT.

     

    YTD performance of this basket (with EFA added last month)

     

    1/1 - 4/1 TLT 4.41%
    4/1 - 5/1 IEV 4.01%
    5/1- 6/1 EEM -5.05%
    6/1 - 7/1 IEV -2.29%
    7/1 - 7/31 QQQ 3.73%

     

    YTD Net 4.5%

     

    Indices YTD
    QQQ 8.7%
    SPY 3.35
    DIA 0.53%

     

    Selection for August for close of market on 8/3/2015: QQQ

     

    I own QQQ.
    1 Aug 2015, 01:16 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    Varan:

     

    Thanks for the update!
    1 Aug 2015, 01:20 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » If I may be excused for the self-serving reference, this might be of general interest.

     

    http://bit.ly/1DZFbGL
    1 Aug 2015, 04:05 PM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    Good article Varan. My best success thus far with models is when I simply follow the model without tinkering.
    1 Aug 2015, 06:39 PM Reply Like
  • rabsparks
    , contributor
    Comments (31) | Send Message
     
    Varan: If you've run any backtests on your modified model (ijj, qqq, eem, epp, iev, efa, and tlt), I would appreciate knowing how it performed in the past.
    5 Aug 2015, 02:05 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » 2003-todate

     

    CAGR 25%
    Max DD 14.6%
    Sharpe 1.29
    Sortino 2.74
    Worst Year 2004 (10% gain) excl 2015
    2015 return 4.1%

     

    Hope this helps.
    5 Aug 2015, 03:09 PM Reply Like
  • rabsparks
    , contributor
    Comments (31) | Send Message
     
    Thanks a lot, Varan. It certainly helps.

     

    Regards.
    5 Aug 2015, 05:40 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @varan,

     

    I calculate somewhat different results:

     

    START/END DATES: 07/31/2003 - 07/31/2015 (12.0 years)
    Sharpe: 1.30
    CAGR: 27.3%
    Max DD: 18.0%
    MAR ratio: 1.52

     

    Rolling 12-month returns:
    Mean: 28.2%
    Median: 24.4%
    Range: -9.5% to +98.0%
    95% CI Limits: -5.7% to +62.2%

     

    Annual Returns:
    2003: 28.30% (= 81.87% annualized)
    2004: 16.92%
    2005: 18.92%
    2006: 27.99%
    2007: 30.57%
    2008: 26.30%
    2009: 55.74%
    2010: 20.39%
    2011: 35.77%
    2012: 20.55%
    2013: 27.23%
    2014: 17.56%
    2015: 4.61% (YTD)

     

    TMD
    6 Aug 2015, 08:22 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @varan,

     

    Do you calculate Max DD on a MONTHLY or DAILY basis?

     

    Thanks,

     

    TMD
    7 Aug 2015, 02:53 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » monthly.
    7 Aug 2015, 06:18 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Ahh....I think that accounts for the differences.

     

    PortfolioVisualizer also reports monthly Max DD, and their figures match up pretty well with your own. They report a Max DD of -14.41%, which is pretty close to your figure (-14.6%). Their Sharpe matches yours pretty closely as well (1.29 vs. 1.31), but there's a bit of discordance in the Sortino ratios (2.74 vs. 2.54).

     

    All in all, I'd have to say PortfolioVisualizer results corroborate your results reasonably well.

     

    TMD
    7 Aug 2015, 10:18 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Interesting.

     

    My returns are based on the transaction made on the first trading day of the month. I do not know if that is causing this difference.

     

    My data source is Yahoo Finance

     

    Thanks.
    6 Aug 2015, 09:13 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @varan,

     

    I also used Yahoo Finance. But I use the last trading day of the month. All transactions take place MOC on the last trading day of each month. The look-back is 3 calendar months, not 63 sessions. I know that can make a difference as well.

     

    PortfolioVisualizer yields figures closer to yours than to mine:

     

    (I tried to post the link, but SA truncated it so it won't work)

     

    You calculate a MAR ratio of 1.71, I make that 1.52, and PortfolioVisualizer has it at 1.78. I'd take any of those numbers personally. All are excellent.

     

    I think we can conclude that during this test period at least, the strategy produces annual returns of about 25% with drawdowns substantially less than that (roughly 14% to 18%). Either way, that calculates out to an outstanding MAR ratio that is well over 1.0.

     

    In my book, if you can achieve an overall MAR greater than 1.5 with a double-digit CAGR, you really have something outstanding, especially if the CAGR is > 20%.

     

    Now, that doesn't tell you anything about how much annualized CAGR, Max DD, Sharpe, and MAR ratios vary. But I have analyzed that as well, and I can tell you that these spend most of the time in a reasonably constrained range, which I think is a plus.

     

    TMD
    6 Aug 2015, 11:38 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Thanks.

     

    Actually that's what keeps me very skeptical. Anything that gives you more than 12% with low drawdown is good enough for me. Too much more, above 15%, is too good to be true to last a long time. I will continue to invest in this and track it on these pages to see how it goes.
    7 Aug 2015, 01:00 AM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    Talking about what's good enough for anyone, Varan, do you know of a strategy that improves on SPY buy-and-hold but only trades SPY (or SPY and SH)?
    7 Aug 2015, 07:55 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » that sounds interesting. I have no idea, Do you have any links?

     

    thanks.
    7 Aug 2015, 08:17 PM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    My question wasn't rhetorical. I meant it literally.
    7 Aug 2015, 09:31 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @Algyros,

     

    Maybe SPY with a 10 month moving average stop to cash, assessed only at the end of the month? So, on the last day of the month, assess where SPY is relative to its 10 month moving average. If above, hold (or buy), if below, sell and go to cash. Do not assess intra-month (to minimize whipsaws).

     

    (This is what Mebane Faber does with the Ivy Portfolio, which markedly improves the Max DD but preserves the CAGR).

     

    TMD
    7 Aug 2015, 10:21 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @Algyros,

     

    Looks like SPY with a 12-month SMA works pretty well (better than a 10-month SMA) over the last 20 years:

     

    CAGR 12.23%
    Max DD: -16.63%
    Sharpe: 0.94
    Sortino: 1.54
    MAR: 0.74
    (but 3 years when there were small losses: 2000 [-0.95%], 2010 [-1.13%], and 2011 [-1.90%]).

     

    In contrast, for SPY held throughout these 20 years:
    CAGR: 9.73%
    Max DD: -50.95%
    Sharpe: 0.53
    Sortino: 0.76
    MAR: 0.19
    (4 losing years, including -37% in 2008)

     

    TMD
    7 Aug 2015, 11:22 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @varan,

     

    Looks like there might be problems with the GMR strategy.

     

    Cliff Smith identified the following proxies for the GMR ETFs that allow backtesting to the beginning of 2000:

     

    IJJ--MDY
    EFA--FSIIX
    IEV--VEURX
    EPP--VPACX
    EEM--VEIEX
    QQQ--QQQ
    TLT--VUSTX

     

    Using these proxies, the results were not good:

     

    Returns for 2000 and 2001 were both negative (-13.2% and -28.5% respectively).

     

    The proxies gave comparable results to the ETFs for the period 2004 through present.

     

    We believe that the bear market of 2000 - 2003 is critically important for any backtest. This was a true bear market. Arguably, nothing since truly qualifies as a typical bear market. We have seen this phenomena before: a strategy might produce outstanding performance metrics after 2003, but when you test back through the 2000 - 2003 bear market and/or into the 1990s, you see a very different picture. I strongly suspect that in many (but not all) cases this is due to the strong skewing of strategy performance to the upside that began immediately after QE was instituted in December 2008.

     

    TMD
    12 Aug 2015, 06:11 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Originally I had done a longer term backtest with the equivalent mutual funds FDVLX, FIEUX, FEMKX, FLATX, FPBFX, VUSTX for period starting in 1991, and the results of that basket are pretty good (19.8% 1991-, worst years 1992 -1.3%, 1994 -24.8% 2000 -5.5% and 2008 -6.2%). That convinced me that this was worthwhile to follow.

     

    With the current basket, the FDVLX, QQQ, FIEUX. FEMKX, FPBFX, FWWFX, VUSTX, the 2000- performance is not bad either (I have used FWWFX as a proxy for EFA which is not quite correct, but the two have the correlation of 94%). CAGR 17.7%, worst years 2000: -4.8%, 2002 -2.2%, 2008 -4.8%).

     

    You can perform a longer term calculation by using the index ^IXIC as a proxy for QQQ. So FDVLX, ^IXIC, FIEUX. FEMKX, FPBFX, FWWFX, VUSTX yield the following for 1991-: CAGR 19.2% Worst Years: 1994: -16% 1997: -.05%,2000 -9.9%, 2008 -5.1%.

     

    Not as steller as the ETF basket, but pretty good long term performance, and nowhere as apocalyptic as might appear on the basis of other tests.

     

    Of course one cannot implement the mutual funds strategy as holding periods of one month or lower are a no-no, but the calculations do give some comfort.
    12 Aug 2015, 09:28 PM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    If you add a 9 month moving average filter, you avoid the worst of 2000-2003 with returns of:

     

    2000: -13.17
    2001: -.22
    2002: 1.46

     

    Max DD: -29.27

     

    CAGR: 16.28

     

    Not nearly as impressive, but still four times better than buy and hold SPY. Portfolio Visualizer won't let me do a 10 month moving average filter without jumping to results starting in 2001.
    12 Aug 2015, 09:35 PM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    Thanks Varan. No system is perfect, but this is one of the best I've seen.
    12 Aug 2015, 10:25 PM Reply Like
  • Rt Sentiment
    , contributor
    Comments (23) | Send Message
     
    A better proxy for QQQ is ^NDX.
    15 Aug 2015, 12:13 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @Rt Sentiment

     

    Thanks.

     

    I have half a mind to say that I just wanted to err on the side of caution and that's why I used ^IXIC :)

     

    With ^NDX the results are even better:

     

    1991:2015 CAGR 21.5% Max DD 23% Worst Years: 1994 -15.9% 2000:-4.9% 2002 -1.3% 2008 -4.8% Sharpe .99 Sortino 2
    15 Aug 2015, 12:49 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @varan,

     

    Using FDVLX, FIEUX, FEMKX, FLATX, FPBFX, VUSTX with a 3 month look-back period and no cash default, I get:

     

    CAGR: 17.7%
    Max DD (monthly): -26.5%
    MAR Ratio: 0.67
    Sharpe Ratio: 0.90
    Sortino Ratio: 1.52

     

    Three losing years:

     

    1994: -21.0%
    2000: -3.9%
    2008: -4.7%

     

    Also, FLATX began on 04/19/1993, so the test begins with calendar year 1994. I don't know how you tested back to 1991; I presume you just deleted FLATX for the years 1991 - 1994.

     

    I tried deleting FLATX, and that improved the CAGR to 19.7%, the Sharpe to 0.99, and the Sortino to 1.76, but that came at the cost of a worse Max DD: -30.4%.

     

    It appears that much of the outperformance of the strategy (with FLATX) was due to recent (i.e., post-QE) returns. From 1994 through 2000, the strategy performed poorly:

     

    CAGR: 10.9%
    Max DD: -21.0%
    MAR Ratio: 0.52
    Sharpe Ratio: 0.40
    Sortino Ratio: 0.58
    Two losing years: 1994 (-21.0%) and 2000 (-3.9%)

     

    By way of comparison, the SPX performed much better:

     

    CAGR: 18.3%
    Max DD: -15.4%
    MAR Ratio: 1.19
    Sharpe Ratio: 0.90
    Sortino Ratio: 1.41

     

    You can slice and dice performance according to various start and end dates, and you will find a lot of variability in performance. This underscores my contention that any backtest that is analyzed using just one set of start/end dates provides very limited insight into how the strategy performs in differing economic climates, and how it might perform going forward.

     

    I maintain that static metrics like CAGR, Sharpe, Sortino, etc. give just a freeze-frame view, and are highly susceptible to start/end date bias.

     

    To me, the only reasonable approach if one hopes to acquire some modicum of a comprehensive insight into a strategy's performance is to analyze distributions of standard metrics, for example by using rolling return periods. Anything less will give very limited insight into how a strategy really performs.

     

    TMD
    16 Aug 2015, 01:27 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    Varan or TMD:

     

    I am tracking nine indexes using 3 month relative strength. I go long the top 2 or 3 and go short the bottom 1 or 2. The short sales must be in a downtrend before I will short them.

     

    Have either of you done any work on Long/Short momentum?

     

    I am currently long QLD and SSO. I am short EEM and DBC.
    7 Aug 2015, 10:27 AM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @extremebanker,

     

    That's an intriguing idea. I would like to know more. What is the basket of funds?

     

    I have played around with something conceptually similar, but simpler (and cheaper) to execute: use pairs of long/short ETPs. So all positions would technically be long, and short positions would be via inverse ETPs.

     

    I haven't found anything that really performs well so far, but I haven't tried very hard either. Applying the trend filter would be easier with this approach because you would apply the same criteria to all ETPs. But that said, not all ETPs are created equal, and while something like a 10 month moving average might work well in terms of defining a long-term trend for some ETPs, it would probably not work for others (e.g., treasuries, currencies, some commodities).

     

    TMD
    7 Aug 2015, 03:03 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    TMD:
    The basket of funds is SSO,QLD,EFA,EEM,VNQ

     

    RSP,IJR,DBC and TLT

     

    I have run some long only strategies with SSO and SH with reasonable success. But high volatility. SPY and SH does not seem to work well.

     

    The leveraged products allow me to have excess cash to short other positions without borrowing money. I don't go long or short in equal manner. Currently, I am 95% long and 25% short. I can't run simulations on ETF replay and just thought someone may have done some work in this area.
    7 Aug 2015, 03:37 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    TMD: I have several objectives with this strategy.

     

    Generate absolute returns. Just performing better than the market does not suit me. If the market goes down 40% and I only go down 20% that is good relative performance but not what I am looking for.

     

    Reduce volatility with short positions. Good short positions can reduce volatility below market levels.

     

    Try to benefit from asset rotation rather than asset allocation. The funds covered work with domestic stocks, small cap stocks, international stocks, emerging market stocks, real estate, commodities and bonds. That is a lot of different asset classes. I also track Euro and Yen futures as a separate part of this strategy.

     

    Try to take advantage of some leverage with this strategy. I try to think of each position as a "long" or "short". I rank each fund weekly to help answer this question. I look at the relative strength of each fund and what the MA position is. I also use stock index futures with this strategy. ESH16 and NQZ15. Try to take advantage of backwardation, no management fees and no interest charges for leverage.

     

    I have performed better than SPY for the year but not by a large amount. I am tracking my performance and want this to work better than average in all markets. We will see.
    7 Aug 2015, 04:40 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @extremebanker,

     

    Just using a straight momentum strategy, I couldn't find any set of parameters that gave a Sharpe greater than 1.0 or a Max DD any better than -28%.

     

    (I realize this is just a very crude approximation of what you are doing.)

     

    In situations like this when there are variable amounts of volatility, a risk parity approach might be helpful.

     

    TMD
    7 Aug 2015, 11:32 PM Reply Like
  • Algyros
    , contributor
    Comments (164) | Send Message
     
    Extremebanker,
    Could you give more details on your strategy: what instruments are you using and any performance data you might have? Thanks.
    7 Aug 2015, 12:03 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    It is similar to what is being tracked here. I can back test long only on ETF replay. I have been selling short the bottom of the momentum ladder for only a short while. I can't back test the short sales on ETF replay. Just wondered if anyone could back test the difference in the strategy.
    7 Aug 2015, 12:32 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    Recently Monte Carlo Analysis was introduced in my backtest platform of choice. While I'm still exploring this feature and its implications, so bear with me, using pretty much out-of-the-box settings, Monte Carlo Analysis may reveal real risk hiding in the darker corners of the strategies we discuss over here.

     

    The Monte Carlo simulation was run for the new composition IJJ, EFA, IEV, EPP, QQQ, EEM, TLT.
    First, a backtest starting December 31, 2003 and ending July 31, 2015 is performed. The backtest follows the GMR Simple rules for monthly portfolio reformation, but with trades executed at MOC.
    Next, from this original set of 139 trades the Monte Carlo engine randomly picks trades to produce a new, random set of again 139 trades. Some original trades may be skipped and some may be used more than once (permutation with repetition, or random sampling with replacement). This trade randomization process is repeated many times, for the below results even 1,000,000 times.
    Finally, during post-process distribution statistics and charts are generated from the data obtained during the randomization process.

     

    The equity curve and statistics from the initial backtest look familiar to the results reported by Varan and others (me too):
    http://bit.ly/1Ww83Sg

     

    The outcome of the Monte Carlo Analysis is summarized in a table showing key statistics derived from the distribution of the simulation results:
    http://bit.ly/1Ww85to

     

    One of the charts focuses on MaxDD%:
    http://bit.ly/1Ww83Si

     

    Subject to further investigation, by imposing a reasonably conservative confidence level of 95% it becomes possible to assess the probability of a specific outcome. With the said confidence level in mind, probability states the strategy's MaxDD% will reside somewhere between -11.71% and -100% as reported by the the table and chart (see red window). The reported MaxDD of -14.40% appears to have only a probability of less than 10%...
    16 Aug 2015, 10:35 AM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @TrendXplorer,

     

    Very interesting. Unfortunately, the links to the charts did not work. Maybe you should post them on your blog?

     

    I understand the implications for the Max DD, but what were the results of a similar analysis of CAGR?

     

    Thanks,

     

    TMD
    16 Aug 2015, 03:19 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @Trendxplorer,

     

    The links work now, thanks very much. I don't know what the problem was before, but all's well that ends well.

     

    If I interpret the drawdown graph correctly, the probability of a 30% drawdown is 45%. Is that right?

     

    Thanks,

     

    TMD
    16 Aug 2015, 11:58 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    If the chart is representative - emphasis added on "if" - there is 45% probability that the strategy will have a maxDD of 30% or less.
    The flip side: there is a 55% probability that the strategy's maxDD will be 30% or higher.
    Since we're talking absolute terms, less is better.
    Please do take note of Varan's observation.
    17 Aug 2015, 02:59 AM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    Due to the long tail from -100% the MC CAR graph isn't very informative all, because the charted curve gets too compressed.
    However the MC Final Equity is still presentable:
    http://bit.ly/1PvVdyU

     

    The posted performance chart ends with equity at $1.413K. As the MC Equity distribution graph shows this is just about at the 50th percentile (see also the posted table with key levels).
    17 Aug 2015, 02:48 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I think that the assumption made by Monte Carlo simulation that P(r[m+1] | r[m] ) = P(r[m]), i.e. the distribution of returns for a given month is independent of the return for the prior month leads to this result, and may not be valid for a momentum based strategy such as this.

     

    Actually P(r[m+1]>0|r[m]<0) is around 0.8, so there is some reversal built into the strategy (not by intent, but by the actual result) that is not accounted for by the simulation.

     

    Which is not to say that the Monte Carlo simulation results are not useful, especially for underscoring the element of caution for going into the strategy all in full speed ahead, as I have myself tried to convey in some of my comments.
    16 Aug 2015, 03:31 PM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    Somehow our earlier comments got vaporized...

     

    Albeit the meaning of the "|"-sign is not clear to me, I think you're pinpointing an issue that does bother me.
    The 139 trades have a profit spread from $132K - $-91K. However the distribution of the spread is skewed to the right (see chart below), due to the compounding effect as a result of the strategy's profitability combined with continued full capital commitment.
    Distribution of returns ($): http://bit.ly/1UPh5Ij
    17 Aug 2015, 03:55 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @TrendXplorer,

     

    Interesting chart. Clearly the performance after Dec 2008 is much different than prior to Dec 2008. Monthly returns from 2004 to 12/2008 averaged 1.75%, but after that, they averaged 2.27%.

     

    Also, the win rate and the ratio of profits on winning trades/losses on losing trades was much better post 12/2008:

     

    PRE 12/2008:
    Win Rate: 65%
    Profits on Wins/Losses on Losers: 0.91

     

    POST 12/2008:
    Win rate: 73%
    Profits on Wins/Losses on Losers: 1.70

     

    So both these contribute to the outperformance of the strategy post 12/2008.

     

    TMD
    17 Aug 2015, 07:56 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I am just trying to point out out that the probability distribution of the returns ( for the strategy) for a given month is dependent on the return for the prior month and the simple appplication of Monte Carlo ignores this dependence.

     

    In particular the results show that the probability that the return is positive in a given month given the prior month led to a loss is higher than the probability of positive return in a given month without any precondition. Ignoring this structure on the returns imposed by the strategy may lead to more pessimistic estimates.
    17 Aug 2015, 04:20 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » I did a quick simulation using this (not necessary the perfect) approach to take into account the dependence of a given month's return on the return of the immediately prior month.

     

    (Apologies for going deep into the woods here. I also plead slightly guilty to torturing the results to prove my case. I partially blame TrendXplorer for this, and also thank him.)

     

    A random sequence of 152 monthly returns (covering the period Jan 2003-Aug 2015) was generated as follows:

     

    1. Generate 76 random numbers between 1 and 152.

     

    2. Each of these random numbers is used to index into the sequence of monthly returns of the actual strategy, and a pair of monthly returns is obtained by appending to it the next monthly return from the sequence. (If the index is 152, the first month return is used as the next return in the pair.)

     

    3. These 76 pairs are concatenated to obtain a sequence of 152 monthly returns.

     

    4. This randomly generated sequence of monthly returns is used to compute the performance metrics.

     

    As I suspected, the results are not as pessimistic as those of TrendXplorer's simulation.

     

    (Based on 2000 randomly generated monthly return sequences)

     

    Min. Max DD 8%, Max MaxDD 47%, MaxDD for 95% of the cases under 30%.

     

    Min CAGR 7%. Max CAGR 51% CAGR for 95% of the cases over 15%.

     

    Min Sharpe .46. Max Sharpe 2.4 Sharpe for 95% of the cases over .8.
    21 Aug 2015, 10:28 AM Reply Like
  • TrendXplorer
    , contributor
    Comments (112) | Send Message
     
    Thank you very much for taking the trouble to develop your own Monte Carl engine. Nicely done. And yes, I do take full responsibility, but at the same time I am glad we're deep in the woods now.

     

    I really am, because your results offered a benchmark by proxy for taking another stab at Monte Carlo Analysis. With modified settings, hopefully the right ones this time, my new results pretty much confirm yours. To anyone terrified by my earlier comment, I apologize.

     

    The Monte Carlo methodology is the same as described in my earlier comment, so trades are still taken into account without any dependency with the prior or following month. Statistics are derived from 1,000,000 random simulations again.

     

    The Monte Carlo table holds the key statistics:
    http://bit.ly/1MLxw7w

     

    The MC CAR chart presents the full data range for the compound annual growth rate:
    http://bit.ly/1MLxyw5

     

    The MC Drawdown chart paints MaxDD's distribution:
    http://bit.ly/1MLxyw7

     

    The points of interest are the 95% levels mentioned by Varan:
    - CAGR for 95% of the cases over 15% (match)
    - MaxDD for 95% of the cases under 32% (near match)

     

    The 99% levels are noteworthy too.
    23 Aug 2015, 04:09 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » That's excellent work Sir.

     

    Thanks.
    23 Aug 2015, 10:19 AM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    So are we looking at going to SHY tomorrow?
    31 Aug 2015, 04:35 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @ls78

     

    Yeah it is painful.

     

    Since 5/29/2015 TLT has been the best, but turned in the loss of 0.4%.

     

    So it is SHY or if you are optimistic TLT. I am sticking with TLT.
    31 Aug 2015, 11:27 PM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    I can't imagine there's going to be a rate hike or even expections in that direction anymore for a while. That and more market turbulance should work just fine for TLT.

     

    drftr
    1 Sep 2015, 10:36 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » thanks drftr. that's reassuring. we shall see. as someone says, after putting the head in the oven, it is futile to worry about the heat.
    1 Sep 2015, 11:02 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    Although there is no real global justification for a rate hike I believe the FED is anxious to get the normalization process moving along.

     

    Unemployment is in the low 5% range and GDP is starting to move therefore they may feel comfortable with the initial increases.

     

    Once again, it is a normalization process since rates should have never gone as low as they did.
    1 Sep 2015, 02:52 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @extremebanker,

     

    You might be interested in an article Bill Gross published today on SA. Once you get past all the meandering musings, he paints a rather bleak picture. He basically says that global QE has created a monster. The result will be an environment where any decent returns on investments in either equities or fixed income will be very difficult to produce or sustain. He suggests cash or short-duration fixed income is a viable option.

     

    http://seekingalpha.co...

     

    TMD
    2 Sep 2015, 11:05 PM Reply Like
  • rabsparks
    , contributor
    Comments (31) | Send Message
     
    Maybe we're already seeing the changes that Mr. Gross thinks will happen should the FED up the interest rate this year. But weighting one's portfolio to cash and short duration bonds takes a helluva lot of intestinal fortitude.

     

    And us retired folks, going to relatively liquid investments makes it even harder.
    3 Sep 2015, 12:57 PM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    @rabsparks,

     

    I hear you. Cash just locks in a loss, and short-duration fixed income would at best mitigate the slippage from inflation and taxes.

     

    That seems to echo Gross' feelings also: he thinks we are stuck between a rock and a hard place, and the only reasonable solution is to go back to where we were before QE started.

     

    TMD
    3 Sep 2015, 02:33 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    TMD:

     

    Thanks for the link. Interesting read! I would agree with the general thesis. However, I believe the market could offer better returns with reasonable market timing. I don't think the market will be flat but very choppy which could create decent returns if you are not buy and hold.

     

    I also believe investors need to expand the asset classes they follow to generate reasonable returns. I have been investing in currencies this year looking for better returns. It has been my best performing asset class even though it is a small percentage of my portfolio. I will have to make larger commitments to this asset class going forward.
    4 Sep 2015, 10:02 AM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Thanks EB. I'd like to know more about what you are doing with currencies.

     

    I am more concerned than you about the equities markets going forward. I watched the SPX chop sideways and gradually approach its rising 200 SMA. It was clear something would have to happen when these two met.

     

    Of course, what did happen was a break to the downside that was anything but subtle. The force of this break was, to me, very disturbing. I think it is significant as well, and we are probably looking at some tough months ahead, at the very least.

     

    TMD
    4 Sep 2015, 08:37 PM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    TMD:

     

    My currency trading strategy would probably be called amateurish. I like currencies with positive interest carry. Basically, I trade long term trends. I am currently short the YEN and have been short the EURO several times this year. I don't use leverage since I have more cash on hand than my currency exposure. Cash on hand is very high at the moment and my currency exposure is only about 5% of my portfolio in nominal terms. I intend to increase my portfolio allocation to currencies in the near future. At least 10% of my portfolio. My returns have been in excess of 10% this year. It is just another trade in my view.

     

    I like to track and watch major currencies due to their economic impact on everything. Currency values are as important as interest rates and oil IMO. Trading currencies helps me keep a handle on what is happening in the world.
    5 Sep 2015, 08:22 AM Reply Like
  • tmdoherty
    , contributor
    Comments (1171) | Send Message
     
    Thanks very much extremebanker,

     

    TMD
    6 Sep 2015, 03:05 AM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    extremebanker,

     

    Have you used any of the (inverse) dollar baskets for trading? If so, do you have a preference? I've looked into UDN (using that right now), DBV and ICI but would like to use some other vehicle. In general I don't like the composition and proportions in those ETFs. I know on the futures market there's at least 6 inverse dollar baskets with great diversity. Is there something similar on the mutual fund market that you're aware of?

     

    Thanks!

     

    drftr
    6 Sep 2015, 07:32 AM Reply Like
  • extremebanker
    , contributor
    Comments (2123) | Send Message
     
    drftr:

     

    What I use is the forex market at OX. They clear through FXCM. They pay interest carry and update daily.
    6 Sep 2015, 10:40 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Thanks EB. Good macroanalysis. Hopefully they will punt for a few months again given the reaction of the market.
    1 Sep 2015, 03:24 PM Reply Like
  • 16790612
    , contributor
    Comments (834) | Send Message
     
    Very interesting.

     

    Thank you for this.
    1 Sep 2015, 08:54 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Oct Allocation: TLT
    YTD returns -4.2%

     

    Quite bad, for this is supposed to reduce the DD, but not much worse than the indices:

     

    QQQ -3.44%
    SPY -7,16%
    DIA -8,38%

     

    Someday I should invent a performance metric such that even bad results look good, just as the DGI guys clap at every drop in the value of their portfolios.
    1 Oct 2015, 01:17 AM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    Come on, varan, -4.2% is really not bad this year. Really. As I remember this model got burned bad one month with Latin America (this year? maybe last?). Otherwise, it's done reasonably well.
    1 Oct 2015, 10:52 AM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    Ha ha, thanks for updating Varan. It figures the year I try these strategies is the first one in like 10 where they don't work. Hopefully just a blip. What's your take?

     

    Do you think the roller coaster is due to fears of rising rates, which is throwing everything out of whack?
    1 Oct 2015, 08:58 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » In my view (I am not a trained, or even untrained, finance professional) it's a combination of a number of factors: the age of the bull market, the stage of the presidential election cycle, and of course the interest rate worries. Many of these things cannot be really modeled no matter how much historical data you have, and so the only other option seems to be to get out at the earliest intimations of such hiccups and expect to lose a lot of upside too.

     

    The only strategy (obviously among the ones that I follow) that has worked YTD is the sell in May. If you had bought VTI on 12/31/14 and exchanged it for TLT on 5/31/15, you would have been up over 5% (but this did not do so well in 2013, returning only 10%). The strategy with select funds that I mentioned on my post on the seasonal anomaly some time ago also would have led to about 4% profit.

     

    The utility allocation that I posted early in the year has also avoided the losses for the year, having returned the negligible -0.2% .

     

    The portfolio of AAPL, MSFT, INTC, GOOGL, KR, AZO, COST, NSRGY, UL, SBUX, SRE, WEC, AWR, CELG,
    GILD, AMGN, TLT, IEF, AGG (utilities, tech, biotech, and staples/retail, with around 15% fixed income) has also done OK, having return around 5%.

     

    One good thing that the recent fluctuations have done is to lay to rest the cliche that it is a market of stocks and not the stock market, for I do not think that any of the people who invest on the basis of that would have invested in AMZN or even AZO.
    1 Oct 2015, 10:12 AM Reply Like
  • pat12357
    , contributor
    Comments (257) | Send Message
     
    A few observations:

     

    - "indices" - finally somebody to use the proper plural from of "index"

     

    - "the year I try these strategies is the first one in like 10 where they don't work" - or perhaps it is the first one they are traded live?

     

    - The other day I checked the YTD performance of the logicalinvest (Frank Grossmann) strategies on their website. Looks like they have hard time beating their respective benchmarks (which are not all well chosen, btw) since they have been live.

     

    - For further comparison between backtested and live trading performances one can check portfolio123. Selling shovels remains more profitable than digging for gold...
    1 Oct 2015, 10:32 AM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    Nothing remotely "proper" about it.

     

    "Indexes and indices are both accepted and widely used plurals of the noun index. Both appear throughout the English-speaking world, but indices prevails in varieties of English from outside North America, while indexes is more common in American and Canadian English. " grammarist.com
    1 Oct 2015, 10:44 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Thanks pat. I guess we were both educated in the proper British system, though mine was borrowed.
    1 Oct 2015, 10:39 AM Reply Like
  • pat12357
    , contributor
    Comments (257) | Send Message
     
    The one that really beats me is the use of "haven't" instead of "don't have" when have is an ordinary verb. I see this even in books.

     

    Maybe this too has become accepted and widely used in North America (eh, Left Banker? :) )
    1 Oct 2015, 11:38 AM Reply Like
  • Left Banker
    , contributor
    Comments (3329) | Send Message
     
    Didn't mean to derail here. Apologies. I also did not mean to imply indices was not proper, which I can see is how my comment reads. What I should have said was "nothing remotely improper about indexes."

     

    Fact is, I'm a bit, well more than a bit truth be told, of a grammar geek, albeit less of a prescriptivist than some. I use indexes because indices sounds, to my ear, unecessarily pretentious and pedantic.

     

    We all have out pet peeves on language usage. I try to keep mine to myself ordinarily.
    1 Oct 2015, 11:56 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » October was a bust.

     

    For November 2015, QQQ is the selection.
    2 Nov 2015, 12:06 PM Reply Like
  • ls78
    , contributor
    Comments (180) | Send Message
     
    Looks like QQQ again.
    30 Nov 2015, 05:57 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » By a mile. QQQ is the one.

     

    Thanks.
    30 Nov 2015, 06:03 PM Reply Like
  • tp2
    , contributor
    Comments (42) | Send Message
     
    Varan, I got QQQ, but it did not pass the 3 month moving average. This is for Jan. 2016. Is that correct?
    1 Jan, 11:42 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Without any filter, the selection for Jan 2016 is QQQ again.

     

    The strategy could not recover from the August loss, and ended up with 7.2% loss for the year.

     

    Not much consolation, but the M* fund selector site http://bit.ly/N9TS0j lists GMOKX as the second best allocation fund (select Fund group "Allocation", click for 'show results', and sort by YTD returns) which has a history of only three years (the first has only one year history), and its total return for 2013-2015 is far less (total 6%) than that for the Simple GMR (40%) for the same period.

     

    Thanks.
    1 Jan, 12:24 PM Reply Like
  • tp2
    , contributor
    Comments (42) | Send Message
     
    I got TLT for February. Good luck to all.
    30 Jan, 11:30 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Yes, it is TLT.

     

    Even the last of the guys on CNBC here http://tinyurl.com/jmg... agrees. Good or bad, I don't know.
    31 Jan, 10:41 AM Reply Like
  • ikkyu
    , contributor
    Comments (291) | Send Message
     
    Good article on just how tough 2015 was for finding an asset trend:

     

    http://bit.ly/1PGmGQE
    6 Feb, 03:39 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » That would be March 2016.
    29 Feb, 06:59 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » Sorry, the comment deletion system on SA seems to be messed up.

     

    Please ignore the comment above. I will just report the corrected comment.

     

    For March 2016, the selection is TLT by a very long shot - TLT return for 2015/11/30-2016/2/29 8%, with the next one, EEP having returned -7.4%.
    29 Feb, 07:23 PM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » For April 2016, the selection continues to be TLT.

     

    YTD return -2.2%
    31 Mar, 06:21 PM Reply Like
  • Villi Grdovich
    , contributor
    Comments (904) | Send Message
     
    Any chance of getting a performance graph of how things are going. Bit hard to make sense of the occasional reported number.
    1 Apr, 02:17 AM Reply Like
  • varan
    , contributor
    Comments (5574) | Send Message
     
    Author’s reply » @Villi

     

    http://bit.ly/1VE6525

     

    Is this what you are looking for?
    8 Apr, 10:42 AM Reply Like
  • Villi Grdovich
    , contributor
    Comments (904) | Send Message
     
    Probably. My interest here is with the issue of backtested results not translating into The future. In general, some of this could be due to data snooping and some to behavioral factors or changed conditions ( though it always seems strange to me that conditions change just after a strategy is implemented). From what I see, you have done a lot of backtesting, and now that there is a published record, I was interested to see how this backtested strategy works under the auspices of a person who can be trusted to apply it exactly.
    8 Apr, 10:56 PM Reply Like
  • EdwardjK
    , contributor
    Comments (197) | Send Message
     
    "...though it always seems strange to me that conditions change just after a strategy is implemented."

     

    Villi, your's is not a casual observation. I think the long-term performance of a successful, backtested strategy causes people to ignore the current short-term performance.

     

    I think it important to compare current one-, three- and six-month performance to prior periods to assess whether the strategy holds water.

     

    You have to be wary if current performance does not compare favorably to prior periods.
    19 Apr, 08:45 AM Reply Like
  • drftr
    , contributor
    Comments (1220) | Send Message
     
    "...though it always seems strange to me that conditions change just after a strategy is implemented."

     

    Although I'm starting with the same lines my response would be different:

     

    In the last 2 years many strategies have been invented and described but since we're in a whipsaw kind of period none really diversified strategy has performed the way they did over decades. That doesn't make them all bad, it means that you have to give them more time. Remember results are averages. And yes, you can have a couple of bad years right at the start. But if you have a decent backtested time-frame of at least 15 years and you're reasonably diversified you should be okay. And like Varan has mentioned a couple of times don't expect 20+% strategy returns as an average but as an outlier. If it looks like it's too good to be true it probably is.

     

    When starting using a new strategy always expect to start with the worst years and ask yourself if you would want to stay the course if the first 3 years would be the worst 3 years of the strategy. If the answer is no don't do it!

     

    drftr
    19 Apr, 09:01 AM