Performance

The investment strategy I deploy is a very simple and systematic model. It is based on momentum - the idea that markets with strong relative strength continue outperforming in the short-term while weak markets continue to underperform. You can read more about what momentum is and why it's so pervasive in my first blog post The Philosophy of Momentum.

Specifically, the system I follow is called Global Equities Momentum (GEM). Simply invest 100% in the strongest asset among 3 (US stocks, non-US stocks and bonds) based on past 1-year performance. Details are described by GEM's founder Gary Antonacci in his book Dual Momentum Investing. I encourage you to read it.

The following are my own independently-verified performance results for GEM over the period from Jan 1, 1972 to Mar 31, 2015 (over 43 years).




In the above graph, 
  • RED = US stocks (S&P 500 Index)
  • GREEN = Non-US stocks (MSCI ACWI ex-US Index)
  • PURPLE = Bonds (Barclay's U.S. Aggregate Bond Index)
  • BLUE = Global Equities Momentum (GEM) 

All above results are total return values and stated in Canadian dollars.

GEM is the clear winner. It delivered significantly higher returns compared to both US and non-US stock indices and with lower volatility. This was achieved largely by side-stepping severe bear market drawdowns. 

Over our 43-year test period, GEM made an average of only 1.1 trades per year. It was fully invested in US stocks for approx 40% of the time, non-US stocks for 40% and in bonds for only 20%. Compare this to a traditional portfolio having a fixed allocation to stocks and bonds at all times. With this traditional approach, you suffer in bear markets due to the exposure to stocks and in bull markets due to the exposure to bonds. 

Finally, GEM alleviates concerns of over-optimization and data mining in four ways. First, it has only one simple model parameter: the look-back period (which is set to an elegant, whole-number value of 1 year). Second, momentum has over 2 centuries of backtest data showing persistent outperformance on a decade-by-decade basis. Third, momentum now has several decades of out-of-sample data since first being studied by academia. Fourth, there is a logical explanation for why momentum occurs in markets (please refer to my first blog post).

This is why GEM truly is a gem! Again, be sure to read the Dual Momentum book:



  

68 comments:

  1. What is the source of your data for ACWI ex US prior to 1988? I am a Bloomberg user and cannot find the data of that index during that period.

    ReplyDelete
    Replies
    1. I'm using MSCI World ex-USA prior to 1988. It's basically like ACWI except it's developed countries only. Also, the US Aggregate Bond Index data I have only goes back to 1976. Prior to that, I estimated the index using 5-yr US treasury bond yields. Most data sets (S&P, MSCI, Forex data) only go back to 1970, otherwise I would have tested further back.

      Delete
  2. Gogi,

    I loved your contribution "Dual Momentum for non-US-Investors" on Gary Antonacci's blog. Your piece was well researched and clearly presented. Even though I am a Euro based investor (I know, too much hassle to reconstruct a reference currency for the euro case before 2002), I learned a lot from it.
    I would be very interested to see an overlay of the equity curves for cases 1 through 4. Could you provide the correlation matrix for the equity curves? Furthermore, I would love to see the equity curve of an equal weighting of the strategies of cases 3 and 4. Is there any diversification benefit in terms of reducing maximum drawdown?

    ReplyDelete
    Replies
    1. Hi, thanks for the comment. In my post on Gary's blog, I'm looking for the optimal way to implement GEM, not 4 different variations of GEM to run in parallel for diversification. I can plot an overlay of all cases on one graph (for say, GEM in CAD). I'll do this in a future post when I repost my currency work on this blog.

      Delete
    2. Gogi,

      I knew, of course, what your objective was. I, however, would be interested in whether there is any diversification benefit to running 50% case 3 and 50% case 4 as a strategy. The idea behind this question is to get an impression whether it would be advantageous to run the strategy 50% in an interactive brokers account in case 4 mode and 50% in an account of a European broker in case 3 mode. I am sensing some drawdown reduction might be possible.

      Delete
  3. Gogi,

    I assume you are using 12 month look back period as advocated by Gary. Did you test with any other look back period, say 3 months?

    ReplyDelete
    Replies
    1. Hi Deepinder, yes I'm using a 12-month look back (LB) and monthly rebalancing (RB).

      Having a shorter LB means you're quicker to react to a change in trend but it also means you get more false signals and higher trading costs. Having a longer LB has the opposite effect. A 12-month LB gives the optimal tradeoff.

      Similarly, having a shorter RB will mean you will be more prone to suffer from short-term volatility and get whipsawed. Having a longer RB and you'll be slow to react. A monthly RB is optimal.

      Gary explains this in a recent podcast: http://bettersystemtrader.com/009-gary-antonacci/

      Delete
  4. Gogi, I assume you are using ETFs to implement GEM and generate your own signals. Are you simply switching to a intermediate bond fund such as BND or AGG when equities are underperforming to cash (such as BIL)?

    ReplyDelete
    Replies
    1. Hi Deepinder, you are exactly correct. I'm switching to AGG when both equity ETFs are underperforming BIL on a past 12-month total return basis.

      Delete
  5. This is the only part of the GEM strategy I don't understand. Why not compare the 12-month total return of BND against VOO and VEU and buy the one with highest momentum?
    Thanks for the great work, btw

    ReplyDelete
    Replies
    1. This is the whole idea behind absolute momentum, which looks at how a risky asset (eg. VOO, VEU) has performed over the last 12 months relative to zero (ie. cash). Absolute momentum makes you exit the risky asset after it's own past 12-month performance has turned negative. This is something that all the GEM backtests shown has typically signaled that the asset's uptrend is over and has helped you avoid big drawdowns. It's also why absolute momentum is considered a form of trend following.

      Now, "absolute momentum" is almost the same as "relative momentum between a risky asset and very short-term bonds" (like T-bills, or the BIL ETF). But the advantage of using short-term bonds over cash is that bonds are usually stronger than stocks when stocks are trending down. Hence, using bonds instead of cash means you exit stocks earlier in a new bear market.

      OK, so why not use something like BND (which has a 5-year duration) instead of short-term T-bills (which have just a 3-month duration) like you asked? If we use BND, the we are really just using relative momentum between VOO, VEU, BND and no absolute momentum. But the main reason is that BND's 5-year duration means this ETF is lot more sensitive to interest rates and therefore using it in GEM can cause more whipsaws.

      Delete
  6. Hi Gogi, I've read the book "Dual Momentum Investing" and I found your site because I was looking for Canadian investments that would provide the same benefit. I can see from your study that you also recommend trading the ETF;s on the US Stock exchange. I'm questioning how the conversion from CDN to US dollars at > 25% still makes it worth it? Also, is this a strategy that can be used with RRSP even if you trade on the US Stock exchange? Thanks.

    ReplyDelete
    Replies
    1. Hi Claudio,

      Currency risk was one of my biggest concerns as well. So I spent some time looking at how various non-US investors would have fared in their home currency if they followed GEM over the past 40 years. The conclusion is that GEM inherently protects and benefits from currency fluctuations. How? It's because the relative performance of US to non-US stocks is largely driven by the US Dollar Index. Therefore, when GEM is in US stocks, it is likely that the USD is rising; and when GEM is non-US stocks, it is likely that USD is falling.

      Here's a good example of this in action:

      Since mid-2010 to until recently, GEM has been in US stocks (apart from 4 months). During this time, Canadian followers of GEM got an extra ~30% boost from the USD:CAD gain. From mid-2003 to mid-2008, GEM was in non-US stocks and so Canadians were protected from the ~30% drop in USD:CAD rate.

      Other foreigners (European, Aussie, etc) got similar performance as Canadians. US investors also benefited from currency changes (except they got their boost during 2003-2008 and their protection during 2010-present). This is what I mean by GEM inherently protecting and benefiting from currency changes.

      Delete
    2. I recommend 2 ways for non-US investors to implement GEM. Either:

      1. Stay permanently on your local stock exchange (eg. TSX for Canadians) and use *non-hedged* stock ETFs (eg. VUN, XEF) and a home country bond index (eg. VAB)

      2. Stay on the US exchange when GEM is in stocks and use ETFs such as SPY and ACWX. When GEM is in bonds, transfer to your local exchange and buy a home country bond index (eg. VAB). This is what I'm doing personally. If you go this route, you must use a broker that charges minimal currency exchange fees (Interactive Brokers charges only 0.01% or $2.50, whichever is greater. You can't beat this rate).

      And yes, many Canadian brokerages let you trade in USD in your RRSP.

      Delete
    3. Thank you so much for this. Under options 1 of staying on the TSX, do you use the absolute momentum of US ETF (i.e. VTI), or the Canadian equivalent (i.e. VUN) for the prior 12 months?

      Right now, VTI has very slight absolute momentum (~0.65%), while VUN is 19.5% over the past 12 months due to the exchange rate. Which ETF do you use to gauge your entry/exit decision?

      Delete
    4. Hi Cam. Good question. I looked at what effect it would have if you did the monthly analysis using ETFs priced in USD vs. ETFs priced in your local currency. Long story short, you want to do the analysis using ETFs priced in USD. If you instead did the analysis using ETFs priced in CAD, AUD, GBP or JPY, the past 40+ year annual return of GEM would have fell over 200 basis points on average (ie. you would have earned 16% instead of 18%).

      Look at what I wrote under "Case 2"of my currency study: http://www.dualmomentum.net/2015/06/dual-momentum-for-non-us-investors.html

      Delete
    5. Dear Gogi,
      I have read your article on Antonacci's blog, and I'm really interested (but confused too) about the optimal approach in GEM for non US investors.
      As European investor, it's not easy to understand why I should perform performance analysis in USD instead of local countries currency.
      In principle, local currency analysis would seem the more rational choice. If I use USD (or EUR, it's the same thing) analysis, I add currency performance to equity index performance. At the local currency level, the tracked performance is net of currency fluctuations.
      I am sorry if I miss something, but the question it's not so easy to understand for me.
      Best regards,
      Federico A.

      Delete
  7. Thanks for your detailed explanation.

    By the way since you use a US Exchange when in Equities and a Canadian Exchange when in Bonds doesn't that mean you are converting your money back and forth from one currency to another. Although it may not happen frequently aren't there conversion fee's that you would have to pay? Thanks.

    ReplyDelete
    Replies
    1. As mentioned above - Interactive Brokers charges only 0.01% or $2.50, whichever is greater. You can't beat this rate. Also keep in mind GEM only makes about 1 trade a year and is only in bonds 20% of the time, so you can expect the conversion to only happen once every few years.

      Delete
  8. Are you currently in aggregate bonds at the moment? According to the Global Equities Momentum logic we should be in agg bonds.

    S&P500 (SPY) > ACWX
    T-Bill (BIL) > S&P500 (SPY)

    Thanks.

    ReplyDelete
  9. Based on my calculations GEM should be in SPY in September. 12 month return till end of August SPY (0.36) > BIL (-0.11) > VEU (-11.77). Allocation could change end of September. It would be good to validate against others.

    ReplyDelete
  10. Hi Deepinder,

    I have re-run the test and came up with the following figures; (as of September 29)

    SPY -4.82% (percent price Change)
    ACWX -16.00%
    Bill -0.11%

    Therefore, is SPY > ACWX = YES
    is SPY > BIL = NO

    GEM = Aggregate bonds.

    If i'm incorrect please let me know where as I do not see it.

    Thanks.

    ReplyDelete
    Replies
    1. My understanding is that calculations are done end of the month to determine position for next month. Based on end of August calculations (SPY (0.36) > BIL (-0.11) > VEU (-11.77)) September position should be SPY. Based on September end calculations, October position is likely to be bonds as your numbers above show. I also look at performance numbers on Gary's website and mine match from Jan to August. September isn't complete so I can't tell what is his current position.

      Delete
    2. In both mine and Gary's backtests, we both used end-of-month values since that's how historical data is stated. In implementing GEM, it's perfectly fine to do the analysis any day of the month as long as you stick to that same day.

      Delete
  11. By the way gentlemen may I ask what website or software you are using to obtain your data.

    Thanks.

    ReplyDelete
    Replies
    1. I am downloading monthly data from yahoo finance using a python script I wrote. I analyze and switch position (if needed) end of the month.

      Delete
    2. Careful with using data from Yahoo finance. I did my backtest with Yahoo data and got results that were significantly lower than Gary's.

      The cleanest source of data is straight from the index providers:
      1. Standard and Poor's for S&P500 (Go to http://ca.spindices.com/indices/equity/sp-500 and click table view above the chart. "Total Return" is what you want)
      2. MSCI for All-Country ex US Index (Go to https://www.msci.com/end-of-day-data-search and select "All-Country" for market, "Gross" for Index level and "Standard" for size).
      3. US Fed for T-bills (Go to http://www.federalreserve.gov/releases/h15/data.htm and scroll down to Treasury Bills, Secondary Market -> 3-month)

      I know it's tedious going to all these sites each month. The next best thing is to use StockCharts.com. You can bookmark this link and visit it once a month: http://stockcharts.com/h-sc/ui?s=SPY&p=D&yr=1&mn=0&dy=0&id=p41039941477&a=407337590

      Delete
    3. Hi Gogi, Came across your blog from Antonacci's tweet. I've tried replicating GEM strategy myself with Yahoo Finance data. I backtested from May-1997 to Aug-2015 using SPY, CWI + VGTSX, LAG + VBMFX and 1-month USD Libor for risk free rate. My results were close enough to Gary's in this period (13.31% GEM v/s 12.40% mine) Data from index providers does not account for fees and expenses so one has to be careful in taking those results at face value.

      Delete
  12. Hello Gogi,

    If I want to stay in the CAD ETF market what do you think of:

    HXS for US market (no distribution so you save on Wh tax)
    80%XEF and 20%XEC to replace :ACWX (you could also add a 5% of HXT for canada)

    and ZAG for bonds

    Cheers.

    Christian

    ReplyDelete
    Replies
    1. HXS is interesting - I didn't know we had an ETF that track S&P TR. However, it looks like this ETF hedges currency rate (the past 1yr performance of HXS matches S&P 500 TR in USD). You do not want to hedge currencies as I've shown here: http://www.dualmomentum.net/2015/06/dual-momentum-for-non-us-investors.html

      ZAG is a good bond choice, as is VAB. See chart below:
      http://stockcharts.com/h-sc/ui?s=VAB.TO&p=W&st=2012-01-02&en=today&id=p01943159464&a=421876551

      Delete
    2. I don't think HXS is hedged -- I could be wrong but I downloaded their fact sheet from the horizon's site and it says No. Another benefit of HXS over SPY is you avoid US Estate tax (if your account is that large). Too bad the Liberals killed the corporate class switches -- I was using Purpose ETFs and BMO ETF Corp Class funds.

      Delete
    3. Sorry one more benefit -- there are no distributions on HXS so "dividends" are not taxed as they use a total return swap....whereas dividends from VUN will be taxed.

      Delete
  13. One of the comments that you left above was extremely helpful. I have a few more questions though.

    Q1. To down load Standard and Poor's for S&P500, I went to http://ca.spindices.com/indices/equity/sp-500, and click table view above the chart. "Total Return" column was shown, but I was not able to download the historical data.

    Q2. I downloaded MSCI World ex USA between 1974-1988, and MSCI ACWI ex USA between 1998-2013. I calculated monthly returns, and their mean and stdev. However, when I compared my numbers to Table 8.1 of Gary's book. I found some difference.
    My number: mean = 10%, stdev = 17%
    Gary's number: mean = 8.51%, stdeb = 17.65%

    3. You also mentioned US Fed for T-bills. How did you use this data?
    I think the purple line in your graph above is for Barclay's AGG total return index, which is not equal to US Fed T-bill yields.

    ReplyDelete
    Replies
    1. 1. On that S&P link, click on "additional info" button on the left then scroll down to "monthly and annual returns." A spreadsheet will download with data back to 1988. S&P no longer gives away the data going back to 1970

      2. Make sure you select "gross" for index level and "standard" for market cap. Also, refer to Gary's recent blog posts for performance - I think he found an error since writing his book and updated the #s (they improved).

      Delete
    2. 3. T-bills are short-term bonds (3-month yields). BIL is a bond ETF containing T-bills. If the past 12-month performance of BIL is greater than SPY and ACWX, then you want to be invested in AGG (Barclays Aggregate Bonds) which are medium-term bonds (~5-year yields). The purple line in my graph above is AGG from 1976-present. Prior to that, I had to approximate AGG by downloading 5-year yields from the US Fed Reserve website. This site explains how you can calculate a total return bond index from just yields: http://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/histretSP.html

      Delete
  14. Hi Gogi,
    How important do you feel it is to use Total Return data to do the analysis that generates the signals. Gary uses Total Return but the index providers you mentioned don't.
    Do you think its a good enough proxy without Total Return ?
    Thanks
    Will

    ReplyDelete
  15. Hi Gogi,
    How important do you feel it is to use Total Return data to do the analysis that generates the signals. Gary uses Total Return but the index providers you mentioned don't.
    Do you think its a good enough proxy without Total Return ?
    Thanks
    Will

    ReplyDelete
    Replies
    1. Hi Will. You must use total return data for both backtesting and for ongoing monthly analysis.
      For backtesting, I got my data from the index providers (MSCI and S&P) and both give you total return data.
      For monthly analysis, use StockCharts.com (PerfCharts feature) instead of Google Finance.

      Delete
  16. This comment has been removed by the author.

    ReplyDelete
  17. Great work Gogi! I noticed that in GEM, they used ex-US. Though here on the Canadian stock exchange, you had recommended XEF which excludes Canadian stocks. Do you think that makes a difference? I was thinking about using either VIU or VDU and was leaning towards VIU. Would do this inside an RRSP.

    Also, although deviating from GEM, do you think that adding VCN to the mix as a third alternative to VUN/VIU when in equities (eg) when VCN>VTI/VEU might improve returns?

    Thanks.

    ReplyDelete
    Replies
    1. Thanks Rodney. I will get back to you in June.

      Delete
    2. Rodney, sorry for the late reply. It will make a diff whether you use EAFE vs. World ex-US (with the latter being better). Another commenter pointed out below that VDU is now ex-US. You can use that.

      Adding VCN might improve returns - try it an see. Data is available from MSCI

      Delete
  18. Hi,

    I am trying to locate the data for MSCI World ex USA between 1974-1988, and MSCI ACWI ex USA from 1998 onwards.

    Seems like everyone else managed to find this data but for some reason I can't seem to come right.

    Any pointers would be greatly appreciated (or if anyone is willing to share the data they found that would really be great).

    Many thanks

    ReplyDelete
    Replies
    1. Go to MSCI.com -> what we offer -> indexes -> performance. Accept the waiver. Then you should see the database. World ex-US will be under "Developed Markets" and ACWI ex-US will be under "All country"

      Delete
  19. More good news from Vanguard for Canadians -- VDU is changing to all world ex-US as of Dec 15th. If I am not mistaken, this is exactly what is needed under Case 2.

    ReplyDelete
    Replies
    1. Fantastic! Yes, this is what Canadian's needed. An All-World ex-US ETF performs much better than an ETF tracking say EAFE (I showed this in Case 3 in my post on Gary's blog).

      Also, I checked the HXS again, and yes - it is unhedged, my mistake. Definately advantageous from a tax perspective.

      The only thing that worries me about HXS and VDU is the low trading volume and so bid-ask spreads will be high. Keep in mind the performance results shown above don't factor in this liquidity risk. On US exchange, SPY has 1600x the $-volume of HXS and ACWX has 100x the $-volume of VDU. Something to consider...

      I'm studying for next 2 months and won't be able to reply. But thanks again for sharing your findings.

      Delete
    2. I agree re: bid ask spreads -- this is one of the variables you need to be very careful of with illiquid ETFs. Years ago I was trading a preferred share ETF that had spreads of 1-1.5%. It became untenable -- in fact -- it turned out that you'd be better off in an open end mutual fund. Trading 1.5x per year should be workable.

      Delete
    3. HXS spread is about 4c or 0.08% -- with less than 2 trades per year this is manageable. ETFs also have much deeper liquidity than the volume would indicate. Market makers will step in very quickly if there is any arb.

      Delete
    4. Good to know that spreads are low. The other liquidity issue to look at then is market depth. What is the impact on executed price if your trade size is large relative to avg volume?

      Delete
    5. You could hit 10,000 share bids all day long and the market makers will just keep putting up another bid. The liquidity feature of ETFs that allows for creating and redeeming units ensures the mm's will arb any slight difference in price between NAV and trading price. The primary market is all that matters.

      Delete
  20. Wow. What a wonderful blog. Thanks. Read Gary's book a while back but was struggling with implementation/analysis paralysis until coming across your article on his website. I'm a UK based investor with a reasonable sized account who wants to avoid US estate tax problems. Only problem is, while there are plenty of choices for unhedged US equity exposure, there are no ex-US etfs on the London Stock Exchange. Could I cobble together an ACWI equivalent by purchasing the relevant ETF's? Also, to convert GBP to USD now is perhaps not the best time. Any suggestions appreciated?

    ReplyDelete
    Replies
    1. GGrewal has an equally wonderful post on currency for the non-US investor. He is busy studying so here you go: http://www.dualmomentum.net/2015/06/dual-momentum-for-non-us-investors.html Basically you are better off staying in USD and doing your analysis in USD. As for US Estate tax -- in Canada we have a vehicle (HXS on the TSX) that uses a total return swap for the S&P500. Maybe you can search for the same?

      Delete
  21. Amazing Blog. I love it. I just read the book.
    I love your back tests.
    I back tested it as well in Amibroker with dividend adjusted yahoo data. I had about 12 percent per year. Now I am concerned about the data not being total. But Amibroker says it adjusts for dividends using yahoo data. . I am going to try the same back test on the MSCI data as you did.
    My interpretation was that Gary just compares to the T Bill rate. But you use the 12 month ROC of Bil. What exactly is the difference?
    Are you also going into ivv and aswi as Gary did ?

    ReplyDelete
    Replies
    1. I don't trust Yahoo data. You need to use clean data from the index providers themselves (Standard & Poors, MSCI and US Federal Reserve). I will do a post on this - it is the most challenging aspect of backtesting GEM.

      Gary and I both use the 12-month ROC of BIL. I'm using SPY and VEU for equities. You can use Vanguard - doesn't really matter. Sorry for late reply

      Delete
  22. Hi Gogi
    I am implementing GEM in a French tax deferred account on my local stock exchange (Paris - euro). There is indeed a non-hedged SP500 (Amundi ETF S&P 500 UCITS ETF) available for this account, but there is no ACWI Ex-U.S. When GEM ask to be invested in ACWI Ex-U.S I am considering to invest in the four following Amundi ETF MSCI ETFs that I can trade on this account. I will split the funds respective to their approximate proportion in the ACWI Ex-U.S :

    -48% in Europe ETF (CEU.PA)
    -19% in Japan UCITS ETF (CJ1.PA)
    -12% in Pacific Ex Japan UCITS ETF (CP9.PA)
    -21% in Emerging Markets (AEEM.PA)

    Considering my approximation and the extra cost to invest in 4 instead of just one ETF, do you think this approach as valid?


    I am also willing to start GEM in a local (France) non tax sheltered account so I can invest directly in SPY or ACWX. But should I have the account currency in EURO or US Dollar?

    ReplyDelete
    Replies
    1. Hi WC. You can certainly use those 4 ETFs when GEM is in World ex-US, though it seems very cumbersome. I don't know too much about European exchanges and what ETFs they offer. If they don't have an World ex-US or EAFE etf, then maybe use just Europe. Do a backtest and see what happens - its the best way to learn.

      For second question, I would think having the account in US dollars would make sense if you're trading US-listed stocks. Sorry for the late reply

      Delete
  23. Hi Gogi,
    CASE 4: Non-Currency Hedging the Aggregate Bond Index: Appears to give the best return for AUD based investors by 0.03% p.a. Is there any reason why Australian investors wouldn't invest directly into the US market? Apart from the perhaps fact it may be easier to use a Vangaurd ETF's (which now include World Ex-US) on the Australian Stock Exchange for equities and bonds (hedged or unhedged).

    ReplyDelete
    Replies
    1. All foreign investors can be permanently on the US exchange. GEM's performance isn't significantly affected whether you're in US bonds vs. your home country bonds.

      Delete
  24. 1. (sorry, I didn't read Antonacci's book) How long has it been since Gary published his book and how did GEM performed since then?
    2. Instead of having US and non-US stocks as possible alternatives, adding more specific regions in that mix (like US, EU, Japan, Emerging Markets etc), should increase the performance, no? At least, theoretically. Did anyone run some tests like this? And if the performance is worse, wouldn't it tell us we may do some overfitting in current GEM?
    3. If I am not wrong, a sort of GEM with a larger mix of alternatives is implemented in http://logical-invest.com/portfolio-items/global-market-rotation-strategy/. Its performance in 'out of sample'/live performance is not impressive, imho. To the extend that the idea behind that is similar to GEM, its weak performance can raise some questions on GEM's validity?

    ReplyDelete
    Replies
    1. 1. You can find GEM performance on Gary's blog: http://www.optimalmomentum.com/gem_trackrecord.html

      2. I get asked this a lot. Using just US and non-US is simple, making it less prone to overfitting and overtrading. It also benefits from macro trends in the US dollar (for more, see: http://www.sharpereturns.ca/2015/07/markets-are-interconnected.html). I looked at adding EEM to the model, results are in comments section here: http://www.sharpereturns.ca/2015/07/how-would-gem-adapt-if-emerging-markets.html

      3. Every active strategy will inevitably have a period where it is weak and underperforms its benchmark. That doesn't make the strategy invalid. Complexity, insufficient data and lack of explanation are what make a strategy invalid.

      Delete
  25. This comment has been removed by the author.

    ReplyDelete
    Replies
    1. The most important thing about rebalancing is that you pick a day of the month and stick to it. That said, the day you pick will have some impact on your future returns, but this risk is entirely random. Corey @ Newfound Blog calls this timing risk. To reduce it, he partitions the portfolio and uses different rebalance dates on each partition. I don't think this is necessary - it results in higher trading costs. You can read more here:

      https://blog.thinknewfound.com/2015/08/prepared-market-breakout-breakdown/
      https://blog.thinknewfound.com/2015/04/new-research-paper-minimizing-timing-luck-portfolio-tranching/

      Delete
  26. Hmm... sounds complicated and this should be simple. For the back testing did you use the last day of every month?

    ReplyDelete
    Replies
    1. Pick a day, stick to it. Simple.

      Delete
    2. There is research out there that proves that some days are better than others. Our clients rebalance on a day that gives them a significant advantage over using a random static day.

      Delete
    3. Yes there is a statistical advantage to using a specific day.

      Delete