SystemTrader // Introduction // Setting Ground Rules // Picking the Securities // Buy-and-Hold Performance // Reducing Drawdowns with Simple Timing // Increasing Returns with Pure Momentum // Combining Trend-Following and Momentum // Conclusions - Past Performance.... // Asset Group ETFs // Further Study //// ........
Introduction
Today we will dive into the world of asset class rotation by testing some trading systems using 15 securities. I will start by showing how a basic buy-and-hold strategy would have performed over the last 20 years and then see if we can improve this with a little timing with some trend-following indicators. Timing does improve the drawdowns, but the returns are still a little low so I will then turn to a rotation system based on momentum. Finally, I will combine the two systems, timing and momentum, to show how traders can improve the Compound Annual Return and reduce the Maximum Drawdown.
Revision Note: This article was originally posted on August 12th and was revised on August 15th. The theory remains valid and the returns are still good, but the returns are a little bit lower than originally reported because the system settings inadvertently used the closing prices of the signal for entries instead of the next open. The backtest re-runs have signals on the weekly close and entries on the open of the following week. It is interesting to note that you can often improve returns by taking a trade near the weekly close instead of waiting for next week's open.
I also set the portfolio size at $100,000 instead of $1,000,000. A $10 commission per trade affects a portfolio of one hundred thousand Dollars more than one of one million dollars. However, commissions are not a real concern because the Timing + Momentum system (EMA Crossover and Rate-of-Change Ranking) generated just 233 round trip trades in 19.5 years (233 x 2 x $10 = $4660). This is indeed 4.66% of the initial 100,000, but keep in mind that these commissions are spread over 19.5 years and the portfolio ended with a value of $596,669 (9.60% Compound Annual Return). Thus, commissions accounted for less than 1% of the end value of the portfolio (4660/596669 = .78%).
Setting Ground Rules
First, let's start with ground rules and set tradable universe. The testing period runs from January 1997 until June 2016, which covers almost 20 years (19.5 to be exact). I am using weekly data and signals for this backtest. Signals are based on the weekly close and a position is taken with the next open. I did not account for slippage, but did add $10 commission per trade. The positions in the portfolio are equally weighted and the portfolio size is $100,000. Thus, the position size for a portfolio with 5 positions would be 20% per position.
Picking the Securities
I am using a combination of Vanguard Mutual Funds and Bloomberg Commodity Indexes to create four asset classes and subdivisions within each class. I chose these because of their long data history and the fact that they have corresponding ETFs, a list of which will be shown at the bottom. The year in the first column represents the first full year of data and the numbers preceding the names are used for sorting purposes. There are four bond-related securities, two classified as "other", five equity securities and four commodity securities. I decided not to include currencies. Chartists looking to add a currency element can consider $XAD, $XEU, $XJY and $USD.
1990 VWESX 100061 Bond Vanguard Long-Term Corporate Fund
1990 VWEHX 100063 Bond Vanguard High Yield Corporate Fund
1992 VFISX 100062 Bond Vanguard Short-Term Treasury Fund
1987 VUSTX 100064 Bond Vanguard Long-Term Treasury Fund
1990 VCVSX 100071 Other Vanguard Convertible Securities Fund
1997 VGSIX 100072 Other Vanguard REIT Index Fund
1991 VPACX 100054 Equity Vanguard Pacific Stock Index Fund
1991 VEURX 100053 Equity Vanguard European Stock Index Fund
1995 VEIEX 100055 Equity Vanguard Emerging Markets Stock Index Fund
1993 VDIGX 100056 Equity Vanguard Dividend Growth Income Fund
1987 VFINX 100051 Equity Vanguard Index Fund S&P 500
1991 $BCOMEN 100081 Commodity Bloomberg Energy Subindex
1991 $BCOMIN 100082 Commodity Bloomberg Industrial Metals Subindex
1991 $BCOMPR 100083 Commodity Bloomberg Precious Metals Subindex
1991 $BCOMSO 100084 Commodity Bloomberg Softs Subindex
Here is a comma separated symbol list:
VWESX,VWEHX,VFISX,VUSTX,VCVSX,VGSIX,VPACX,VEURX,VEIEX,VDIGX,VFINX, $BCOMEN,$BCOMIN,$BCOMPR,$BCOMSO
Buy-and-Hold Performance
All but two of the 15 securities gained over the last twenty years. Energy lost over 70% over this period and softs were down over 50%. Soft commodities include sugar, coffee and cotton. US Stocks gained over 250% and long-term US Treasury Bonds gained over 300%. Note that this is total return data that includes payouts. As the chart above showed, there was some serious money to be made in the Vanguard REIT Index Fund (green) over the last twenty years because the mutual fund was up over 600%.
The table below sums up buy-and-hold performance for these 15 securities. Despite some serious gains and an 87% win rate, the Compound Annual Return (CAR) was 5.63% and the Maximum Drawdown was 40%. This means investors would have endured a 40% decline in their portfolio to get a 5.63% Compound Annual Return.
A 40% drawdown is unacceptable for investors and we do not have to look far for culprits. The chart below shows the performance for seven of the securities from 1997 until 2016. Notice how most dipped in 2002-2003 and virtually all asset classes declined during the 2008 financial crisis.
Reducing Drawdowns with Simple Timing
At this point, we need a way to reduce the Maximum Drawdown and increase the return. Let's first start on the drawdown by applying a simple trend following system: the 10-week and 40-week EMAs. Yes, these are the same EMAs used last week and the ones that performed so well for the S&P 500. I even optimized the EMAs for these 15 securities and the 10-40 Week EMA pair was in the top 15 for CAR. In fact, the 11-40 EMA pair produced the best long-only returns.
The chart below shows the Bloomberg Gold Index ($BCOMPR) with a 10-week EMA and 40-week EMA. There were some whipsaws, but all it takes is a few good trends to make up for these whipsaws. Also notice that this trend following system would have kept investors out in 2013, 2014 and 2015.
The table below shows the results for a long-only system and a long-short system. Even though the long-short system produced a higher Compound Annual Return, the Win% was just 38% and the Risk-adjusted Return was more than 2% lower than the long-only system (red oval). I prefer the long-only system. The Maximum Drawdown dropped from 40% to 11.20% (green oval), the exposure dropped to 63% (blue oval) because investors were out during bearish EMA signals and the Win Percent was 10 points higher 48% (black oval). There is still room for improvement.
Increasing Returns with Pure Momentum
The Compound Annual Return was not very good for the EMA system so let's try a momentum system using Rate-of-Change (ROC). This is the purest of all momentum indicators and a ROC ranking provides an instant look at relative performance. The system buys the 5 securities with the highest Rate-of-Change (13-week or 26-week) and these positions are held until the security drops out of the top half. The chart below shows VFINX with the 13-week Rate-of-Change and 26-week Rate-of-Change indicators in 2012.
This system ranks the securities by the 26-week Rate-of-Change, picks the top five performers and holds them until they drop below 7th place. The dropouts are replaced with the security that has highest Rate-of-Change. This is a classic momentum system that buys the strongest performers and bets on the strong getting stronger. Note that there are 15 securities so anything in 7th place or higher is in the top half for performance.
The table above shows all 15 securities sorted by the 26-week Rate-of-Change (6 months). $BCOMSO, VEIEX, VGSIX, VPACX and VFINX are currently the top five performers (above the green line). The blue line marks the cut-off point for the top seven. If bought today, these five would be held until they drop below the blue line (8th place or lower in the ranking). Keep in mind that this table needs to be updated very week. I will show how to create this table using scans in Tuesday's webinar.
The table below sums up the results for the momentum system and the Compound Annual Return improved significantly. Both the 13-week Rate-of-Change and 26-week Rate-of-Change returned over 9% (green ovals). The Win Percents were also above 50% (blue ovals), but the Maximum Drawdowns were above 25% (red ovals), which is still too high. Given the choice, I would choose the 26-week Rate-of-Change because there were fewer trades (471 versus 262) and the Compound Annual Returns were about the same.
Combining Trend Following and Momentum
At this stage we have increased the return with a momentum system (Rate-of-Change) and decreased the Maximum Drawdown with a trend following system (EMAs). Now let's see what happens when we combine the two. This system will work with three steps. First, I will apply a trend condition by requiring the 10-week EMA to be above the 40-week EMA. Securities that meet this requirement will then be ranked by the 26-week Rate-of-Change. I will then pick the top 5 and hold them until they drop below 7th place in the ranking. The table below summarizes the results of this system.
The top line shows the starting point, which is buy-and-hold. The second line shows the results when we combine momentum and trend following. The Compound Annual Return of 9.6% was 71% higher than buy-and-hold (green oval). The Maximum Drawdown was cut from 40% to 16.44% (black oval). The Win Percent was a respectable 55% and the Average Gain was 2.5 times as large as the Average Loss (blue oval). It is not the Holy Grail, but it looks like a pretty robust trading system that chartists can consider for asset rotation.
Conclusions - Past Performance....
The results reflect what happened in the past and this does not guarantee future performance. Nevertheless, it is clear that the combination of momentum and timing can improve returns. The trend-following part of this system means some good trends are needed to be profitable. It also means that there will be plenty of whipsaws and plenty of trades with small losses or gains. Of the 233 trades, the profit-loss ranged from +5% to -5% for 112 trades (48%). There were 46 trades (20%) that gained over 10%. Of these, 26 trades (11%) gained over 20% and there were 7 trades that gained over 40%. Here's the catch. We never know which trade will result in a whipsaw and which one will result in a 40% gain. This is why a system is based on taking every trade and sticking with the system - in sickness and in health....
Asset Group ETFs
For those interested in ETFs, the table below shows the corresponding ETFs with the first full year of trading and the symbol. The numbers in front of the names are used for sorting purposes.
2003 LQD 100001 Bond iShares Investment Grade Corporate ETF
2008 JNK 100002 Bond SPDR Barclays High Yield Bond ETF
2003 SHY 100003 Bond iShares 1-3 Year Treasury Bond ETF
2003 TLT 100004 Bond iShares 20+ Year Treasury Bond ETF
2010 CWB 100005 Other SPDR Barclays Convertible Securities ETF
2001 IYR 100006 Other iShares US Real Estate ETF
2006 VPL 100007 Equity Vanguard FTSE Pacific Stock ETF
2006 VGK 100008 Equity Vanguard FTSE European Stock ETF
1997 EEM 100009 Equity iShares MSCI Emerging Markets Stock ETF
2007 VIG 100010 Equity Vanguard Dividend Appreciation ETF
1987 SPY 100011 Equity SPDR S&P 500 ETF
2007 DBE 100012 Commodity PowerShares DB Energy ETF
2007 DBB 100013 Commodity PowerShares DB Industrial Metals ETF
2007 DBP 100014 Commodity PowerShares DB Precious Metals ETF
2009 JJS 100015 Commodity iPath Bloomberg Softs ETF
LQD,JNK,SHY,TLT,CWB,IYR,VPL,VGK,EEM,VIG,SPY,JJE,JJM,JJP,JJS
Further Study
These trading strategies were tested using Amibroker with the help of Cesar Alvarez of AlvarezQuantTrading.com. For more information about his testing and consulting services, click here.
******************************************************
Other Articles in the SystemTrader Series:
- Introduction to Key Performance Metrics - Quantifying the Golden Cross for the S&P 500
- Golden Cross and 5 Major Indexes - EMAs versus SMAs - Do Short Positions Add Value?
- Short-term Moving Average Crosses - Differences in Major Indexes - Do Trend Filters Work?
- MACD Crosses - Rules to Reduce Drawdowns and Increase Gains - Testing PPO Strategy
Follow me on Twitter @arthurhill - Keep up with my 140 character commentaries.
******************************************************
Thanks for tuning in and have a good day!
--Arthur Hill CMT
Plan your Trade and Trade your Plan
******************************************************
