Seeking Alpha

Vanguard Capital Preservation Strategy: Effect Of Trade Day And Look-Back Period Length

by: Cliff Smith

Further analysis of the Vanguard Capital Preservation (VCP) tactical strategy is presented. The effects of trade day and look-back momentum period on performance and risk are shown.

It is shown that the best trade days are end-of-month (EOM) and first day of the next month (EOM+1). Trading on other days reduces performance and increases risk.

In a parametric study of look-back periods systematically varied from 10 trade days to 30 trade days, it is shown that the 21-day (one calendar month) look-back period is optimal.

The final VCP strategy using a dual momentum approach and backtested to 1988 has a CAGR of 13.0%, a MaxDD of -5.8%, and a MAR of 2.2.

This mutual fund strategy can be traded monthly (every 30 days) on the Vanguard platform without any costs. However, a strict schedule must be followed.

Introduction to Vanguard Capital Preservation Strategy

This article continues the analysis of the Vanguard Capital Preservation [VCP] strategy originally described here. The VCP strategy updates on a monthly schedule and uses a dual momentum approach. In this strategy, there are six Vanguard mutual funds in the basket of funds covering both equity and bond assets, and the two best (highest momentum) funds are selected at the end of each month. The relative strength momentum ranking is based on a one calendar month look-back period. Absolute momentum is used for risk control, i.e. the two funds with the highest relative strength momentum ranking must have returns greater than the money market asset in order to be actually selected. The out-of-market asset is VFIIX (although a money market asset can be used with little decrement in performance).

The basket of funds is the following:

  1. Vanguard Convertible Securities Fund (MUTF:VCVSX)
  2. Vanguard Health Care Fund (MUTF:VGHCX)
  3. Vanguard High Yield Corporate Fund (MUTF:VWEHX)
  4. Vanguard High Yield Tax-Exempt Fund (MUTF:VWAHX)
  5. Vanguard GNMA Fund (MUTF:VFIIX)
  6. Vanguard Intermediate Term Treasury Fund (MUTF:VFITX)

All of these funds have histories that date back to 1986 except VFITX that only goes back to 1991. To backtest to 1988, the Dreyfus U.S. Treasury Intermediate Fund (MUTF:DRGIX) is substituted for VFITX. By backtesting to 1988, the strategy shows that it can successfully handle various market conditions including bull markets and bear markets.

Please take note that a few of the funds presented in this article are slightly different than those described in the previous article. The other change is that the out-of-market asset is now VFIIX instead of a money market asset. These slight changes were made to improve the overall strategy.

Any investor can take the parameters discussed above and insert them into Portfolio Visualizer [PV], a commercially-free backtest software program. PV will backtest the strategy to 1988, plus it will select what funds to select at the end of each month.

Results of VCP Strategy

The backtested results of the VCP strategy are shown below. The backtest results are produced by Portfolio Visualizer [PV]; the timespan is 1988 - present.

Total Return: 1988 - 2015

Annual Returns: 1988 - 2015

Drawdowns: 1988 - 2015 (with S&P 500 included)

Drawdowns: 1988 - 2015 (without S&P 500 included)

Overall Summary: 1988 - 2015

It can be seen that the Compounded Annualized Growth Rate [CAGR] is 13.0%, the Standard Deviation [SD] is 6.7%, and the Maximum Drawdown (MaxDD) is -5.8%. This gives a MAR (CAGR/MaxDD) of 2.24.

How these numbers compare to a buy & hold strategy (rebalanced annually) and the S&P 500 are presented in the table above. For the buy & hold strategy, the CAGR is 9.0%, the SD is 5.3%, and the MaxDD is -14.3%. This gives a MAR of 0.63. Thus, the tactical strategy is a significant upgrade to the buy & hold strategy. Likewise, the tactical strategy is significantly better that the S&P 500 that has a CAGR of 10.4%, a SD of 14.5%, a MaxDD of -51.0%, and a MAR of 0.20. There are no negative years for the VCP strategy; the worst year has a positive 1.9% return (in 2002). This compares with a worst year of negative 37.0% for the S&P 500 (in 2008) and a worst year of negative 10.3% (in 2008) for the buy & hold strategy.

Further Assessment of VCP Strategy

In this article, further analysis of the VCP strategy will be presented. In particular, the effect of trade day on backtest results will be assessed, as will the effect of look-back period length.

Herbert Haynes has developed a backtester that can be used to study these effects. Haynes' backtester using dual momentum was set up a little different that the dual momentum approach by PV. In particular, the absolute momentum part of the Haynes' backtester is slightly different than PV's absolute momentum test. Haynes followed the conventional absolute momentum technique by Gary Antonacci that uses pure cash or any other asset as the absolute momentum test, and then uses that same asset as the out-of-market asset. In PV, the absolute momentum test is always money market (i.e. 1-month T-Bill returns), and the out-of-market asset can be anything specified by the user.

So for the VCP strategy using PV, the absolute momentum test was money market, and the out-of-market asset was VFIIX. For the Haynes' backtester, the absolute momentum test was VFIIX, and the out-of-market asset was VFIIX. This slight variation between calculations did not cause any significant difference between PV results and Haynes' backtester results for EOM calculations.

First Parametric Study: Trade Day vs. Number of Assets

Using the Haynes' backtester, we first looked at the effect of trade day on performance and risk. For this parametric study, we independently varied the number of assets selected each month (1, 2, and 3) and the trade day. The trade day was varied between EOM-10 trade days and EOM+10 trade days.

Heatmap results are shown below. They were skillfully created by Herbert Haynes. Heatmaps are presented for CAGR, MaxDD, and Sharpe Ratio. The colors range from red being worst to blue being best. So cold spots [blue] are desired for each variable. The numbers on the top of each heatmap (-10 to 10) correspond to the trade day. Zero (not actually specified) corresponds to the EOM. The number [-1] stands for EOM-1. The number [1] signified EOM+1. The numbers on the left (1 to 3) correspond to the number of assets selected each month in the VCP strategy.

CAGR: Range = 8.5% [red] to 15.3% [blue]

MaxDD: Range = -27.5% [red] to -6.8% [blue]

Sharpe Ratio (CAGR/SD): Range = 0.85 [red] to 2.04 [blue]

The heatmaps show that the best trading days center around EOM-1 to EOM+1. The optimal number of assets seems to be two when both CAGR and MaxDD are considered.

Second Parametric Study: Trade Day vs. Look-back Length

The number of assets was set to two, and another parametric was run on Haynes' backtester. In this parametric study, trade day and look-back length were independently varied. The results are shown below in the form of heatmaps. Heatmaps are presented for CAGR, MaxDD, Volatility (Standard Deviation), and MAR (CAGR/MaxDD). The numbers on the top of each heatmap are the trade days as previously discussed, and the numbers to the left of each heatmap are the look-back trade days for the relative strength momentum. The look-back trade days range from 10 days to 30 days.

CAGR: Range = 8.2% [red] to 14.1% [blue]

MaxDD: Range = -27.3% [red] to -6.3% [blue]

Volatility [SD]: Range = 6.3% [red] to 8.3% [blue]

MAR [CAGR/MaxDD]: Range = 0.3 [red] to 2.1 [blue]

For CAGR, an optimum band is seen going from the upper left corner to the lower right corner. Short look-back periods (11 to 14 days) combined with trading between EOM-8 to EOM-1 seem to be optimal and robust. But the MaxDD results show a different optimal window: look-back periods between 20 - 23 days and trade days between EOM and EOM+2. In terms of volatility, a vertical optimal band is seen that occurs between EOM and EOM+2. The MAR heatmap shows an optimal window between look-back periods of 20 days and 26 days, and trade days between EOM and EOM+2.

Overall, the optimal window seems to be around one-month in look-back length, and EOM and EOM+1 in trade days.


  1. The analysis presented in this article indicates that two assets should be selected in the VCP strategy (from a basket of six assets).
  2. The analysis also indicates that the VCP strategy should be traded at EOM or EOM+1. Trading on other days may significantly reduce returns and increase drawdown.
  3. The optimal momentum look-back period is one calendar month.

Some Practical Issues

After further study, it now seems that trading mutual funds on a monthly schedule can only be accomplished using the same family of mutual funds. When different families of funds are used in a monthly strategy, sell and buy trades cannot be executed on the same day. This prevents the execution of a monthly tactical strategy using mutual funds if funds from different families are used.

This issue is circumvented when the basket of funds are all in the same family. Then you can sell and buy funds on the same day. That is why only Vanguard funds are used in the actual application of this strategy.

This is important because Vanguard blocks the buying of a fund for 30 calendar days after the fund has been redeemed. But this 30-day trade restriction can be accommodated in a monthly schedule if the trade day moves around slightly between EOM and EOM+1. I have presented a trading schedule in my previous article that will satisfy the 30-day trading restriction. It must be followed rigorously, or the trade day will slip downstream. And, as shown, trading on days other than EOM or EOM+1 reduces return and increases risk.

The only drawback in this application is that selections must sometimes be made before EOM data are available. In these cases, EOM-1 data must be used to make the selections, with the caveat that there will be some selections that differ from the EOM selections. Going back to 2007, it was seen that EOM-1 selections differed from EOM selections about 17% of the time (averaging 4 selections out of 24 selections each year). This percentage was rather constant over the years. It was also observed that the EOM-1 selections out-performed the EOM selections over the next month about half the time. This seemed to indicate that using EOM-1 data to determine selections is not overly problematic.

It is rather easy to use EOM-1 data to come up with fund selections by using Using PerfCharts, the list of funds is inserted into the symbol box, and the number of days (that varies each month between 20 days and 24 days) is inserted into the slider box. Set the start date at EOM-1 of the preceding month and the end date at EOM-1 of the current month. The percent return is seen to the right in the resulting figure. As an example, the PerfCharts plot for December selections is shown below. The slider box has 21 days for this month. It can be seen that VGHCX and VWAHX are the selections. And please note that they are both greater than absolute momentum, i.e. zero percent return.

We have also found another issue in using EOM PV selections that readers need to be aware of. Many investors will look at PV's selections at EOM and trade accordingly on EOM+1. It turns out that the latest EOM dividend distributions for mutual funds are not usually included in the EOM data feed. This means the adjusted prices are not correct at EOM, and so the selections by PV at EOM may be in error because total returns do not include the latest dividend distribution. The correct adjusted price data are not provided to PV until a number of days after EOM. Thus, the backtest results are correct, but the selections at EOM may be in error using PV.

The only way around this challenge is to calculate total returns yourself by using historical data from a data source such as Yahoo. The Yahoo data will also be in error because the dividend distribution at EOM will not be included. Thus, Yahoo adjusted price data must be modified so that the effect of the latest dividend distribution is included. This is very easy to do and could be automated by skilled Excel users.

Disclosure: I am/we are long VGHCX, VWAHX. 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.