This is the first in a series of tests on rotational systems. Read Part 2.
As described in the jumping-off post, this is a test of the strategy presented in this article: Sector Rotation Strategy: Simple Rotation Trades Just One Fund a Month.
The bottom line is that my results show that the strategy above loses money. I could not get even close to replicating the results. However, my test suffers from some severe limitations (I’m using ETFs instead of the Fidelity Select Funds) which do not allow for exact replication. I’ll address the limitations in a bit.
Here is a good link listing the Fidelity Select Funds (FSFs) with descriptions and their benchmarks: Fidelity Select Fund List.
The Rules:
1) Track the 25 day (or 5 week) price change in all of the Fidelity Select Mutual Funds.
2) Invest in the Fidelity Select Fund with the highest percentage gain over that 5 weeks.
3) Hold that Select fund for at least 30 calendar days, to avoid the Fidelity early redemption fees.
4) After 30 days, if that Select Fund is still the top Select fund, continue to hold it. Otherwise, exchange it immediately for the currently top ranked Select fund.
5) Hold the new Select Fund for 30 calendar days.
The Results:
Click on the graphs to enlarge…
Limitations:
1. I’m backtesting over a list of ETFs which was generously constructed by a reader. View the list here: ETF Equivalents for Fidelity Select Funds. Not only is the list not entirely complete, but not every FSF has an ETF that is highly correlated.
2. Some of the FSFs have been around for almost two decades longer than the ETFs.
3. I set the backtest up quickly, and it is not rotating on exactly the open of the 31st calendar day. Instead, it is rotating on the open of the 2nd trading day of every new month.
4. Other limitations which I haven’t yet discovered.
What’s Next for This System?
I would really like to complete a more accurate test. A quick look at Yahoo data shows that they offer a data history of the FSFs. I should be able to load these up into the ‘ol wayback machine and test over actual FSF data.
Once I get better data, we can attempt some improvements. I’ve already discovered one improvement, although it might not work on the actual FSFs.
I would also like to check and see if I can find more correlated ETFs. I’m sure thankful for Thomas compiling the list, but I should probably double check for better equivalents, just to be sure.
If you would like an Excel sheet with all the trades, email me: Woodshedder73 over at gmail with the subject line of Rotation Strategy Trades.
I loaded up the Fidelity Select data from yahoo and the system is better, but still closer to awful than good.
I’ll post everything soon.
Does anyone happen to know if these funds pay dividends? I’m wondering if there is some dividend adjusted chicanery going on with the yahoo data.
Yes, they do pay dividends. Sometimes BIG ones. The Construction and Housing Select Fund, FSHOX, paid over 4 bucks both in 2006 and 2007! Yahoo historical prices has “adjusted close” numbers that count the dividends and adjust the market price accordingly. You probably want to download and use those numbers. But Yahoo is notorious for having bad data. So be careful. Looking at FSHOX historical data, it looks like the FSHOX dividend isn’t being counted right.
And also with FSHOX, good luck trying to find a comparable ETF. There are actually two, ITB and PKB, but when charting the two against FSHOX, FSHOX falls right in the middle for a 3.5 year chart.
Update…When using Morningstar’s “growth of $10,000” charts, (which includes dividends), it turns out that PKB is a much closer ETF to FSHOX than is ITB.
FSDCX in December 2000 paid a whopping dividend of $13.70 on a $40 NAV.
Since you were kind enough to link to one of my posts, let me make a couple of suggestions as you look for a way to trade these funds. Let me note that I’m not sure of who first popularized the single fund Fidelity system, but Mark Pankin published it for a long time.
I personally wouldn’t trade the system outlined in the particular post that you are linking to. The drawdowns and volatility are too much for me. The system that we’ve published since Feb of 2002 is shown below
http://fundztrader.com/best-fidelity-select-funds/fidelity-select-trading-system/
The key changes made from the system you were looking at are:
1) Penalize volatility: Basically, instead of just using the recent price change as the figure of merit, use the price change divided by the standard deviation of the recent daily price changes ( measured as a percentage).
2) If a fund’s figure of merit is not above a small positive threshold, go to cash and don’t take a trade until a fund exceeds that threshold.
3) Trade 2 funds instead of 1. I’ve actually tested it up to 8 funds, but after 4 the performance starts to tail off, but it tends to be a little less volatile and has a little better UPI than a one fund version.
There’s a little more detail in the series of articles starting with this one
http://fundztrader.com/blog/2007/05/22/fund-selection/
Recently we started publishing an ETF system that basically uses similar rules, but it only has about 6 months of real time trading history. My experience is that ETF’s don’t have quite the price persistence of the Fidelity Select Funds, so it’s a little harder to construct a system that matches the FIdelity performance, which is why we haven’t published an ETF system all these years. We’ll see how this works out over time.
http://fundztrader.com/etf-trading-system/2-fund-etf-trading-system/
Good luck with your ETF system.
P.S. Yes, they pay dividends, that’s why we use data from Fasttrrack because they are pretty good about adjusting for dividends.
John – what is your threshold for getting into a specific fund from money market? I’m assuming it is based on your “ranking” number. Thanks.
champ8x,
I do not think John follows the blog regularly. If you need to contact him, I would go to his site and find his email or a contact form.
I tried. That’s why I hoped this forum might get a response. I think he uses a weighted average compared to a true average (over 20-30 trading days) to determine his rankings, but am not sure how he factors in the volatility.
John, thank you so much for your links and suggestions.
I especially like the idea of requiring a small positive threshold for the figure of merit.
Looking forward to trying out these new (to me) ideas!
Wood,
It looks like most of the damage is done during major bear markets (2001-02, 2007-08). Do you have a fund for cash? During the major downturns cash will outperform the market and be the “top performing fund”. That might reduce the drawdowns and improve overall returns. I suppose this is the same as the “positive figure of merit” theory. Seems to make sense.
I have not included a fund for cash, but I will in the future.
Right now I’m just trying to get AmiBroker to use the adjusted close from yahoo data, to account for the divies.
I can already tell you that a moving average filter is going to help a lot.