My recent posts on the Fidelity Sector Fund Rotational Strategy generated many comments. The following strategy was suggested by the commenter named Redshark.
The problem with these rotational strategies is that unless someone is giving out the signals or the investor has the ability to set up the strategy in Excel, there is no easy way for the investor to trade the rotational system. There are simply too many variables to calculate by hand, and most investors do not have the time or inclination to learn R or learn how to code in Tradestation or the like.
What I like about Redshark’s idea is that as the title states, it is very easy to calculate the signals. In fact, all one needs is the most basic of charting packages.
I have tested it over the Fidelity Sector Funds, which I particularly like because they can be traded with no slippage and zero commissions. This makes them the perfect candidate for testing over as historical results are more likely to be able to be generalized into the future.
All that being said, here are the rules:
- Buy the 3 funds that have been above their 50 day moving averages for the most days
- Hold the funds for at least 30 days
- On the 30th day, if the $SPX is beneath its 50 day moving average, all funds will be liquidated the next day and the system will not trade again until the $SPX is above its 50 day moving average.
- On the 30th day, if the open positions are still in the top 3, do nothing and re-evaluate the next day OR if a fund(s) is not still ranked in the top 3, sell it on the 31st day and buy the fund that has replaced it in the top 3.
That is all there is to it. I have not at all optimized the variables. I chose 50 days for both because I simply prefer the 50 day moving average.
Results from 1.1.2000 to 2.2.2012
- Compound Annual Return: 12.21%
- Winners: 60.78%
- Maximum System Drawdown: -32.10%
- Sharpe: 0.74
Equity Curve:
Drawdowns:
Historical Profit Table:
I expect that most readers will wonder what this will do with ETFs. I’m wondering too.
Redshark suggested using bonds. I did add two Fidelity bonds (FBNDX and FTBFX) to the portfolio, and they had a deleterious effect on performance.
So there you have it: A very simple rotational system that beats buy-and-hold with reduced drawdowns, and you don’t need any specialized software to generate the signals.
WHOOPS: I did not turn off the moving average filter when I added the bonds. Since the idea is to have the bonds step in to replace the funds when the market starts to fall, the test with the bonds was probably flawed. I’ll re-run it with bonds tomorrow and the moving average filter and report back tomorrow evening.
Pretty good idea, Redshark. Kudos to Wood for the programming and data.
And I’ll comment on ETFs in Excel. So far, a strategy of picking the top three ETFs (out of nine) based on geometric returns over the previous 30 to 120 days, with no holding period (i.e., daily switching), and staying fully invested, underperforms SPY from 1998 (ETF inception) to 2012. And you get many one day switches.
The good thing about Excel is there are tons of statistics you can see regarding how the system is working. The bad thing is I can’t even imagine how to implement a minimum holding period.
Bozo, avoid the one day switches. You want to catch momentum. Momo will come and go but over all will persist over longer time frames. I’d start with a minimum of a 5 day hold. Also, commissions will kill you with one day switches.
Keep in mind that there is an inverse relationship between amount of time between switching and commissions. If it costs -0.10% in performance to hold a trade a few days longer, you have to consider the commissions that you would have incurred had you instead closed that position and opened a new one.
There are a lot of tradeoffs to consider, but it is my humble opinion that the longer you can hold the trades, the better.
I completely agree, but there’s a problem: when I say I can’t imagine how to implement a holding period, I really mean it – I have no clue how to write an equation in Excel to hold for longer than overnight. It should be simple but it’s not. I don’t even know what to search for. I probably need a couple of macros, and my skills are weak there. Unless a reader has a suggestion, for now I’m stuck.
My best advice would be to set it to only trade the first day of the month or the last day of the month. You should be able to build some simple logic formulas using the EOMONTH() function and MATCH() to get you where you need to be pretty quickly. I think you should be able to do this easily without VB.
Yes, thanks for the suggestion. I can try that. I had thought about only allowing trades every two weeks, but that eliminates a lot of potential trades that might be worthwhile. Right now I am adding some columns to count how many days we have been in a trade, which can override the other columns that would cause a switch.
Hi Woodshedder
I like the idea, but a max drawdown of 32,10% ?? How many investors are keen on this ?
I know it was meant to be simple, but in my opinion not too many investors are willing to accept this drawdown.
Best regards from Denmark
Lars, I hear you. 32% is not optimal, but it is a little more than half of the max drawdown of the S&P over the same time period. Relatively speaking, that is good. Couple half the drawdown and trouncing the S&P returns, not too bad for simple, eh?
And even better Drawndown vs the S&P in 2008. I think you have to consider as Woodshedder stated that the S&P was down over 46% in that time period and over 38% in 2008, plus Woodshedder more than delivered on the upside. I think the CAGR for the SP 500 over the same period in the backtest is about 1%.
Hi Woodshedder !
I agree with you on that, but that requires investors to stick with the system even through the toughest of times. That may prove quite difficult to the worst enemy of ours: emotions.
No doubt. I know from first hand experience that it is hard.
I’m guessing adding the Bond funds in will help with drawdown and boost returns a little because of the yield on the funds???
Hmmm. I’ll know in a few minutes.
Adding the two bond funds I mentioned in the post and removing the moving average filter reduces CAR to 4.14% and increases the max system drawdown to -53.15%.
With the moving average filter, performance decreases to ~8% CAR with worse drawdowns.
Hey Wood, check out this ->http://quantingdutchman.wordpress.com/2010/10/24/my-thoughts-on-go-in-cash/
I guess Frank over at Engineering Returns had written some code that let’s you go into Bond ETFs in lieu of cash. This guy didn’t post the code, but he says it is available if you email him, and I know you have contacted Frank in the past. I think this is the direction everyone is thinking about. Also, you can see how his performance was increased using this idea. Best wishes, and thanks to you for running this for us.
Thanks for the idea Red! I’ll take a look at that post over the weekend. The Dutchman and I have been in contact a time or two and Frank is a wonderfully generous fellow.
Yes! This is what I was thinking. Using SHY or similar to rotate into when sector fund/etf was undesirable. Thanks for finding Red.
Switching into a bond fund when it outperforms ETF’s in some “lazy portfolios” during a ranking period (quarterly for this paper) doubles the return of the lazy portfolios from 2003-2011. http://ssrn.com/abstract=1917044
Thanks kidd, I’ll take a look at it!
I’ve looked through earlier posts and cannot find any discussion about why you went with this system instead of Mebane Farber’s 5, 10 or 20 ETF strategy. I don’t have his book handy, but as I recall the CAGR is similar but Farber’s drawdown is significantly lower than yours above. I’m pretty sure the ETFs could be recreated with Fidelity sectors. Thanks.
Craftsman, there was really no choice or reason as to why I tried this system with Fidelity funds. Mainly I guess I used it because I had the skeleton of the code already existing from a previous project.
I don’t have Mebane’s book but I have read his papers numerous times. Without violating a copyright, if you could give me more info as to what you are referring to, i.e. the 5,10,20 ETF strategy, I would love to take a look at it. I do think I’ll go ahead and order his book. I assume you are referring to the Ivy League Portfolio? Thanks.
“The Ivy Portfolio” has a section on 20 ETF portfolios, and a section on rotation strategies, but not both combined, as far as I can tell.
Regarding the 20 ETFs, which is a variation on several Lazy Man’s portfolios (no rotation), the data shown in the book is for 1985 to 2008, so it’s not really comparable to the system being discussed here. If it’s extended to 2012 the returns and drawdown (12% and -2%) would be much worse than they had been in the earlier period due to higher correlations.
For the rotation strategy, he buys the top 1, 2, or 3 performing asset classes every month based on rolling 3, 6, and 12 month returns. The drawdowns are all on the order of 30%, from 1973 to 2008.
Bozo, from what I could tell, my Fidelity rotation strategy generated returns and drawdowns that seemed to be similar to Faber’s sector rotational model.
I did test out using rolling 3, 6, and 12 month returns, but an improvement was not easily discernible.
Woodshedder, yes. One of his Ivy portfolios calls for breaking a portfolio into 5 categories, S&P500, MSCI EAFE, 10 year Treasury, commodities (GSCI) and real estate (NAREIT). The model goes long when the category is above the 10 month SMA. If it drops below the 10 month SMA go to cash. In the 25 year period from 1973-2008. the CAGR was 11.33%, with a max drawdown of 9.53%
I’ve studied extensively the data/research behind this as published in his paper Global Tactical Asset Allocation.
Redshark, a blog friend and commenter has noted that he cannot replicate these results using ETFs.
I have not tried it myself.
This rotational system is not using differing asset classes but rather U.S. market sectors. There is quite a difference. I’m corresponding with another fellow who is seeking to combine some sector rotation with asset class rotation.
Do you know what results Redshark came up with?
Let me ask a question real quick if you guys do not mind. I know it might sound dumb, but I am perplexed by it. The 10 Month SMA is just the 200 Day SMA, no? If not Wood, how do you code that in Amibroker?
Well Red, that is a good question. Some months have 21 trading days. A good approximation is the 200 day SMA. However, it is better to use TimeFrameSet(inMonthly);
That way you are using an actual monthly moving average.
Here is some example code. I’m not an expert at it. Just learning it.
SPY=Foreign(“SPY”, “C”);
TimeFrameSet(inMonthly);
monthlyC = C;
MAMonthly = MA(monthlyC, 10);
TimeFrameRestore();
RestorePriceArrays();
The code below I’m not sure of. Kind of making it up as I go along.
Buy = IIf(SPY > TimeFrameExpand(MAMonthly, inMonthly), 1, 0);
Let me know if it works Red. I’ve got about a dozen irons in the fire right now, including learning how to use that timeframeset function.
No need for SMA or anything like that. Every 13 weeks starting from the close of the first full week of the year, invest equally in VNQ, DBC, VTI, and EFA, except that if any one of them performed worse than TLT (long term treasury ETF) during the period of twelve weeks ending at the close of previous week, invest in TLT instead.
During 2007-2011, this leads to the CAGR of 22% without a single year of loss.
I think that performance of an equity with respect to another asset that is negatively correlated with the equity is a much better and more reliable signal than SMA based signals. In the case above, TLT is ngatively correlated with VTI, etc. and that is why you get such good result. Blind relative strength based trading as recommended by Faber is not as good as this approach, and I suspect that explains the reason behind the failure of GTAA so far.
Gregor, do you have any statistics on your suggested fix going back say 20 years. I’m not set up to do backtesting. Thanks.
Craftsman, see my link above in my comment re: replacing equity portfolios with TLT. Results from 2003. http://ssrn.com/abstract=1917044
Gregor, I’ve tested something similar to that, written about here: http://seekingalpha.com/instablog/709762-varan/251242-a-low-drawdown-strategy-for-sector-rotation-for-fidelity-select-funds
My results are no where NEAR what he is getting. My results show a system that loses money or is flat across that time period.
However, rotating into TLT from 2008-2011 did well, but I believe that is an anomaly and is unlikely to be repeated. I still need to do more work on that code but will likely blog about it soon.
Wood, took a look at your linked strategy. How much of the time did it spend in Treasuries? Reason I ask is that the 20 year time frame consists of a big bull market for TLT, so if it spent much of the time in TLT (or equivalent) the results may be once-in-a-lifetime. Also, most academic momentum studies seem to focus on 3, 6 or 12 month. You picked 2 month which is unusual. Did you backtest the system with longer momentum qualifying periods?
Mebane Faber’s ETF, GTAA, apparently implementing his own strategy described in quite popular papers, has not done very well- it generated a loss of 7.3% in 2011, and a gain of 3.02% in 2012. Not much to write about.
No, it hasn’t done very well. I was so impressed with his work, I bought 100 shares a few months after inception. They have not done well at all. In fact, they have underperformed everything else I’ve held.
Woodshedder,
Very interesting analysis!
But it appears that there is a “short-term trading fee” that Fidelity charges if these funds are held for less than 60 days. From their literature, this fee is a $75.00 flat fee, assuming an online trade, even though the fine print at the bottom of the funds table you linked to indicates a short-term trading fee of “up to .75%”.
Has your model taken into account the effect of this short-term fee for funds that are held for less than 60 days? Or have I misinterpreted Fidelity’s short-term fee policy? Thanks.
RG
RG, the fidelity sector funds have a short-term redemption fee of .75% if they are held for fewer than 30 days. They can be sold without penalty on the 31st day. Some of their funds have longer short-term redemption fees, but the sector funds are 30 days. ALL of their funds have a roundtrip penalty which can result in being locked out of a fund or even your account if it is violated. The roundtrip is 30 days. So basically, no matter what, you are selling no earlier than the 31st day.