Screen Spotlight: GLD,RSP,TLT,ZIV

Screen ChartWith this post we begin a new series of spotlighting the screens that show interesting performance on the new Risk Return Chart we introduced last week.

The first screen in the spotlight is  “GLD,RSP,TLT,ZIV”,  which is appropriately named because those are the only four symbols included in the screen.  The screen backtests well and looks promising on the Risk Return chart, but that is not the only reason it was chosen.  We often receive emails about how to interpret screens.  This screen breaks many rules on the surface, but in the end is a  very simple screen and may be effective.  Let’s begin with a look at the Risk Return chart and see where this screen stacks up.  For this evaluationRisk Return Chart we were interested in 1-month holding periods only and were looking for lower volatility.  We began by changing a couple of the default settings.  First, a 1-month re-balancing period was selected, and then the maximum Standard Deviation was set to 30.  Most of us would cringe at a 30% std dev of our portfolio.

In this chart you will see the backtests for this screen are shown in yellow.  This is a new feature of the Risk Return Chart that was released today, allowing you to highlight the family of related backtests.  As you evaluate this family you should note they are clustered in the top left of all the backtests shown, indicating that for the risk taken they provide a good return (in the backtests, at least). That is what first drew our attention to this screen, so on with the evaluation.

Looking at the family on the Risk Return Chart one would expect fewer positions to have a higher volatility and usually a higher CAGR than models with more positions held.  In this case that is generally what happens.  The 1-posiiton backtest is the far right hand yellow dot at about 19% s.d.  The 2-position model is at 14% s.d. and has a higher CAGR which is not totally unusual since the top position sometimes under-performs on momentum screens.  The three and four position models are inside a 10% s.d. and have lower CAGRs than the 2-position model, and the 5-position model has the lowest s.d. and CAGR for this family.  Of course, there are only 4 symbols available to this model so the 5-position would always have 20% cash plus the 4 available symbols, assuming all four passed the filters. In summary, the backtests here indicate a rational screen.

Since there are only four symbols in this screen let’s focus on the 1 and 2 position backtests.  The chart forBacktest chart the 2-position chart is shown above and can be compared to the single position chart shown here.  The chart above is much smoother with drawdowns typically half when holding two positions rather than one.  For this reason, we will focus on the 2-position backest from here forward.

The first thing we generally look at is the screen definition, and we discount results from any screen that appears to be overly complex.  At first glance this screen would appear to be such a screen since there is an entry in every line.  Upon further evaluation, however, you find that most of these entries are meaningless with fields set to != 0.  You also notice that _UV2 is not used in the screen in any meaningful fashion so it can be discounted as well.  The only lines that are meaningful in this screen are the _UV1 definition and the sort rule, and of course the symbol list.  To prove this theory another backtest was run with all the excess removed, and the results were the same.

An interesting detail of this screen is the risk-adjusted return defined in _UV1 and used in the sort rule.  Where many screens use [Rtn-3mo] for a sort field, and Premium Access users utilize the RANK function to risk adjust return, this screen simply divides the 3-month return by the average true range.  There is nothing optimal about that, but a quick test indicates it can be an improvement over the 3-month return alone.

Another quick check recommended for evaluating screens is to look in the comments section to see when the screen was first published, and then take a look at the chart to judge how the screen has performed since that time.  In this case, the screen was published only 7 months ago, which isn’t much time, but has performed reasonably well since then.

The one big question about this screen is the inclusion of ZIV, one of the many VIX related ETNs. VIX can be a great trading tool no doubt, but does it belong in a monthly traded screen?    If you have an opinion on this we would welcome a comment in the section below.

What are your thoughts or opinions on this screen?  What screen would you like to see spotlighted?  Leave a comment.

 

 

 

9 thoughts on “Screen Spotlight: GLD,RSP,TLT,ZIV

  1. Interesting article, and it demonstrates the power of the new Risk Return Chart very well. As to the inclusion of ZIV, I personally don’t have a problem with it, since I regularly trade options against its much more volatile cousin, SVXY. In fact, replacing ZIV with SVXY (of course) improves the CAGR of the screen, albeit with a corresponding increase in volatility:

    http://www.etfscreen.com/screener.php?sbt=641d4

    Removing ZIV from the screen highlights its contribution to the CAGR, and reduces the overall CAGR to about the same as SPY, but with a lower Ulcer Index:

    http://www.etfscreen.com/screener.php?sbt=49f68

    Of course, all of the volatility ETFs can (and will) experience severe drawdowns at times, so limiting losses with stops or puts should be considered, and options are not available on ZIV.

  2. Hugh,
    I also saw some of the rules were redundant. The one that surprised
    me most was changing [TRet-3mo] !=0 to simply >0 reduces the CAGR and
    increases the UI a bit. It then dawned on me that, if both this and the ATR
    are negative, the ratio in the user variable will be positive, probably leading to the improvement noted. All the screens were improved using EDV instead of TLT. One can also replace ZIV by CSD and get pretty good results, at least lately.
    http://www.etfscreen.com/screener.php?sbt=669b6

    A while ago you told me where to find your updated histories of the sector and country screen experiments you have, but I’ve misplaced the URLs. If you still keep those records I would appreciate the URLs for them.

    rrjjgg

  3. Immediately after posting, it dawned on me the quarterly return rule I mentioned could be deleted, so I tried a few more experiments and was surprised that keeping
    the best 3 out of the 4 choices still beat the CAGR of SPY but had really nice
    UI and StDev stats:
    http://www.etfscreen.com/screener.php?sbt=669bb
    That suggests a method for the volatility challenged like my spouse(?).
    Even better, for low UI and StDev it seems to make a good ‘lazy’ portfolio, without too much loss to SPY.
    http://www.etfscreen.com/screener.php?sbt=669bd

  4. Very interesting post. Regarding ZIV, I have been looking into ways to more safely trade the volatility ETFs. I have found through back testing that a good way to reduce significant drawdowns is to get out of ZIV when f1-f2 contango on the VIX goes into backwardation. I don’t know if there is a way to include that in the etfscreen backtests.

    • That is an interesting observation regarding the VIX going into backwardation. Unfortunately I’m not sure we can fit that into the site in the short term. – hmTodd

  5. An interesting screen. For the adventurous, substitute UBT,SSO,and UGL with XIV for a double leveraged screen. Because of the inherent nature of the leveraged funds, choosing the top 2 for 10 days results in a CAGR of 48.8 with a surprising (to me) low UI of 5.6. Both screens leave one guessing as to the impact of a 10% to 15% correction or a sustained downturn in the market. The 5 year backtest has been a one-way street. (Add a market timing rule?)

    • A suggestion for testing the screens in downturn market- reverse the market data from the last five years and you will receive long therm “bear” market conditions.

Leave a Reply

Your email address will not be published. Required fields are marked *