Risk Management Through Cluster Weighting Momentum Analysis

Risk Management is a critical part of portfolio management as it eventually plays directly into portfolio return.  Beginning with a stable of “critical” ETFs, the following Cluster Weighting Momentum (CWM) analysis identifies the favored securities based on data as of 2/12/2014.  ETFs within this list are found in nearly all the portfolios tracked here at ITA.

ETF Rankings:  The positive market at the end of last week is still reflected in the high number of ETFs that are outperforming SHY.  Note that VSS just took over the top spot and all “systems” are go as it has a clean “green” record across the EMA spectrum.  The absolute acceleration or momentum percentage is positive.

I also want to draw attention to VNQ as it continues to move up the list.  I picked up shares of VNQ in several portfolios yesterday.

ETF Rankings

Cluster Identification:  This cluster diagram is on the small size as I needed to crunch it down so my screen-capture software was able to corral it in one shot.  I used a 0.80 correlation cutoff and this ended up breaking the portfolio into nine clusters.  Set a little higher and VWO would have been split into a separate cluster.  Since it is the poorest performing ETF it would not have made any difference.

Cluster #7 (Gold) is the only one to fall short of SHY.

Cluster Diagram

Buy-Hold-Sell Recommendations:  The favored ETFs are: SHY, VUG, VSS, VNQ, LQD, BWX, and PCY.  TLT is a strong competitor to LQD.  Based on the following recommendations for a $100,000 portfolio, only VSS shows up with a recommendation exceeding 20%.

BHS

Cluster Rankings:  The following table isolates the favored ETFs based on the most recent information.  I managed to pick up shares of VNQ, DBC, and PCY in at least one portfolio yesterday.

Cluster Rankings

Comments

  1. David Bernat says

    Lower,

    Let me ask you a question, how is it possible that the models is recommending SHY, my understanding was that SHY was only recommended when all ETF were performing below it.

    Bests

    • says

      David,

      Perhaps HedgeHunter will add a comment, but I think this was due to how I set the filter. Note that I set it at SHY. I just checked and setting the filter at 18 or below removes any percentage being allocated to SHY. Setting the filter at 19 or 20 will result in a small percentage being allocated to SHY.

      Lowell

  2. David Bernat says

    Thanks for your comment Lowell

    Related to the filter, I don’t understand how it is used, can you please provide me with detailed information or redirect me to a link when I can find information about how to use the filter? can I use it to filter a max of 4 assets for example?

    thanks

    • says

      David,

      The “Required Percentage” calculation is proprietary and was developed by HedgeHunter.

      When you are running your own analysis you can set the filter to kick out four ETFs, assuming there are four that qualify. Currently many more than four qualify. If I set the filter to 8, only three show up. However, if I set the screen to 9, the 9th ranked ETF, JNK, will show up as it both qualifies for purchase and it is ranked Number 9 in the rankings. If I set the screen to 14, both LQD and PCY are recommended as LQD is ranked 10 and PCY holds the # 14 ranking spot.

      Lowell

  3. HedgeHunter says

    David,

    In Lowell’s first screenshot above that includes all 30 ETFs in his “critical” list SHY has a “weighted” value of 20.50. Note that, in this full listing DBC has a “weighted” value of 19.4 i.e. lower than 20.50. If we were not using the Cluster Analysis with a SHY cut-off there would therefore be an allocation of funds to DBC but not to SHY.

    However, after reducing the list to 9 ETFs (8 from the SHY filtered cluster list – excluding IAU (Gold) – plus the SHY reference) and re-ranking, note that DBC now has a higher weighting (7.4) than SHY (7.1). The weighting algorithm therefore allocates a small percentage to SHY rather than DBC. This (re-ordering of ranked assets) happens occasionally when the number of assets being ranked is reduced.

    The 2 purposes of the filter (yellow box in Fig 3 above – SHY in this example) are 1) to eliminate assets ranked lower than SHY (by setting SHY as the filter), and 2) to restrict the numsteber of assets (e.g. Max 4) to be included in the portfolio (by adjusting the Filter number until the desired number of ETFs are selected). This results in suggested allocations being generated automatically.

    David

  4. David Bernat says

    Thanks very much for your comments, now it is clear to me

    One more question, what is the “Ranking filter” with two cells “buy” and “sell”. I don’t understand the purpose of that, please can you explain to me or refer me to a link

    thanks very much

    • says

      David,

      While I no longer pay too much attention to the Buy and Sell settings, if you play around with those two options you will see how they impact the “Rank” column. In the above table I have Buy set to 10. ETFs that rank 1 through 9 show up with a Buy in the “Rank” column. The Sell is set to 15 so any ETF ranking higher than 15 shows up as a Sell. Otherwise the ETF shows up as a Hold.

      Lowell

  5. David Bernat says

    Lowell

    Why in the correlation analysis that we do with the Hoadley we never have negative correlations, why is that?

    thanks

    • says

      David,

      The securities we hold never move in opposite directions. It is rare to find ETFs that move opposite to each other unless one is using short instruments. Even bonds tend to move in the same direction as stocks, just not as fast.

      Lowell

    • HedgeHunter says

      David,

      We do get negative correlations if we have the right combination of diversified assets. In my current Hawking Portfolio, UNG (Commodities – Natural Gas), LQD, SCPB and SHV (all Bond ETFs) are all showing negative correlations with my equity ETFs.

      If you are not seeing negative correlations and are looking for a diversified portfolio maybe this is telling you that you’re not looking at the right combination of assets :)

      Great Question

      David

      • HedgeHunter says

        Afterthought Comment:

        If you are referring to the correlation numbers in the Cluster Hierarchy Diagram rather than the correlation matrix itself this is because these numbers only give us information on the strength of the correlation – not the direction (positive or negative). e.g. SPY and SH will show a correlation of 1 i.e. it is strong – but it does not tell us that it is a negative correlation. This is not a problem since this will become obvious from the rankings.

        I have commented on this apparent anomaly in other Posts but you may have missed it.

        David

  6. David Bernat says

    Lowell

    Yes, I was referring to the correlation numbers in the Cluster Hierarchy Diagram, now I see your point.

    thanks