This article presents a volatility targeting scheme that features the Direxion Large Cap Bull 3X Shares ETF (NYSEARCA:SPXL) and the Direxion Daily 30-Year Treasury Bull 3X Shares ETF (NYSEARCA:TMF). I will first look at the scheme using SPXL's and TMF's unleveraged counterparts: the Vanguard 500 Index Fund (MUTF:VFINX) and the Vanguard Long-Term Treasury Fund (MUTF:VUSTX). The Vanguard GNMA Fund (MUTF:VFIIX) is selected as the risk-off asset. This unleveraged strategy adjusts the monthly allocations of VFINX, VUSTX and VFIIX at the end-of-the-month [EOM] depending on their standard deviations over the previous month. The unleveraged version tries to maintain a 4% volatility target. Backtesting the unleveraged scheme to 1987 shows that moderate growth and very low drawdown can be achieved with this strategy.
The strategy is then modified to incorporate 3X leveraged ETFs that have a relatively short lifespan (2009 start dates). The volatility target is increased by a factor of three (to 12%) for the leveraged version of the strategy. From 2010 - present, this leveraged strategy produced a compounded annual growth rate [CAGR] of 21% with a maximum drawdown [MaxDD] based on monthly returns of -10%. This gives a risk adjusted return-to-drawdown ratio MAR (CAGR/MaxDD) greater than 2.
Recent articles by Stanford Chemist on SA and comments by others in those articles encouraged me to try and use my experience with PortfolioVisualizer's volatility targeting scheme to come up with a high performance leveraged tactical strategy. Stanford Chemist wanted to make things simple, using only two assets [the SPDR S&P 500 ETF (NYSEARCA:SPY) and the iShares Barclays 20 Year Treasury Bond Fund ETF (NYSEARCA:TLT)] in a simple relative momentum strategy. The articles showed that a 5-month relative strength lookback period was optimal from 1987 - present using mutual fund substitutes VFINX and VUSTX in place of SPY and TLT. Drftr and others indicated that the optimum look-back period was a function of what decade is studied, and it changes from decade to decade. A 5-month lookback period may or may not produce good performance in the next ten years.
Drftr then challenged us to come up with a momentum scheme that was better than a conventional relative momentum strategy using a single look-back period. To better understand what I am writing about, it may benefit you to look at the articles by Stanford Chemist and the comments to the articles (here, here and here).
As I commented in one of the articles, one of my ideas was to develop a very low drawdown scheme first, and then utilize the same methodology on its leveraged counterparts. One of the issues with developing any type of momentum scheme using leveraged ETFs like SPXL and TMF is the relatively short history of these ETFs. I really don't trust any momentum strategy that can only be backtested to 2009. Harry Long has popularized many leveraged-type strategies like this in SA, but I don't think six years of backtesting is really sufficient. I think a tactical strategy needs to show it is capable of handling market conditions other than just the last seven-year bull market scenario. This is especially true of a strategy using leveraged ETFs that might produce very high drawdowns during bear equity markets (in excess of 75%).
It has been my contention that we need to backtest at least 30 years, and if more data were available, we should backtest 50 years to include market conditions outside the so-called 30-year bull bond market. PortfolioVisualizer is limited to 30 years of backtesting momentum strategies, so I will only go back to 1987 in this article.
Low-Risk Unleveraged Version
I started this study by trying to develop a low volatility targeting scheme using VFINX and VUSTX, with VFIIX as the risk-off asset. I decided to use VFINX and VUSTX as the risk-on assets in a 60%-40% split. The objectives were to achieve moderate annual growth (~8-10%) and very low drawdown (~-5%). Such a scheme then has the potential to produce a CAGR over 20% with 3X leveraged ETFs, and a MaxDD around -15%. The final version of the unleveraged scheme has a 4% volatility target. The PortfolioVisualizer results backtested to 1987 are shown below. The PortfolioVisualizer link with all parameters is shown here.
The strategy is identified as "Timing Portfolio" in the results. It can be seen that CAGR = 8.6%, SD = 5.1%, Worst Year = -0.9%, and MaxDD = -5.8%. The MAR (CAGR/MaxDD) is 1.5. The monthly win rate is 71.5%. The monthly allocation of VFIIX varies from 2.0% to 93.0% over the course of the backtest.
The "Buy & Hold" results are calculated from a 60%-40% split of VFINX-VUSTX, rebalanced annually. The VFINX (S&P 500) results are also shown as a baseline. Both "Buy & Hold" and VFINX show slightly higher CAGRs than this volatility targeting strategy, but their drawdowns are excessively high (-27% and -51% respectively).
High Growth Leveraged Version
Next, the volatility targeting scheme was modified to embrace 3X leveraged ETFs SPXL and TMF. The leveraged ETFs are substituted for VFINX and VUSTX respectively. The only other difference is that the target volatility increased to 12%. Other than these changes, the leveraged version is the same as the unleveraged version.
Backtesting was performed from 2010 - present on both unleveraged and leveraged models. Results from both versions are shown below.
Unleveraged Version Results 2010 - Present
Leveraged Version Results 2010 - Present
The unleveraged version has a CAGR of 8.6% (same as seen in backtest from 1987 to present). The MaxDD is only -4.1%, the worst year is +1.0% (in 2015), and the MAR is 2.1. The 3X leveraged version has a CAGR of 21.0% and a MaxDD of -10.0%. The worst year is -0.8% (in 2015) and the MAR is 2.1. So the leveraged version sees a return roughly 2.5 times the return of the unleveraged version (would expect about three times the return), and a MaxDD that is roughly 2.4 times the MaxDD of the unleveraged version. This shows the overall feasibility of this process, i.e. to develop a low volatility strategy using unleveraged funds, and then substituting leveraged ETFs to increase CAGR while maintaining (or improving MAR).
But please note that backtesting is from 2010 - present for the leveraged version. It is not possible to backtest the leveraged version to 1987. But by backtesting the unleveraged version to 1987, it allows us to argue that the leveraged version will perform well back to 1987 (with a CAGR over 20%) and the MaxDD should be around -15%. These are the numbers we can expect going forward with the leveraged version.
The leveraged version of this strategy is obviously a high risk, high payoff scheme. This scheme should only be used by young investors who desire higher performance with the accompanying higher risk.
In a practical application of the leveraged strategy, the iShares Barclays MBS Fixed-Rate Bond ETF (NYSEARCA:MBB) can be substituted for VFIIX. The leveraged strategy using SPXL, TMF and MBB has returned 17.6% YTD. The allocations for July are 30.4% SPXL, 20.3% TMF, and 49.3% MBB.
Disclosure: I am/we are long 30.4% SPXL, 20.3% TMF, 49.3% MBB.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.