ETF Momentum System

Discussions specific to trading the stock market.
Post Reply
drm7
Roundtable Fellow
Roundtable Fellow
Posts: 96
Joined: Sun Apr 20, 2003 9:02 pm
Location: Richmond, VA

ETF Momentum System

Post by drm7 » Mon Apr 11, 2011 8:21 pm

I spent a lot of time on here and have "taken" a lot of great info from some very generous people. This is my humble effort to "give" a little bit, even though this isn't a huge breakthrough.

The excel 2004 attachment runs a "ranking" momentum system on the sector SDPRs. It's simple. Take the top x performing sectors from last month, buy, and hold them for the next month. Rinse, repeat.

I also added a trend filter - i.e., only take ETF trades this month if the SPY (S&P 500) is above its 10 month EMA last month.

I also added a "short" filter - i.e., apply the trend filter, but also short the SPY if it's below the 10 month EMA.

The conclusion? The ranking system doesn't really beat buying and holding the SPY, but applying the filters definitely does - the "straight" filter gives the best MAR and the "short" filter gives the best CAGR.

Hopefully, you can decipher my spaghetti-like IF statements and tweak it to your liking.
Attachments
ETF System.xls
ETF momentum system
(373.5 KiB) Downloaded 921 times

rgd
Roundtable Knight
Roundtable Knight
Posts: 396
Joined: Thu Aug 10, 2006 5:21 pm

Post by rgd » Tue Apr 12, 2011 12:33 am

Thanks for sharing.

If the ranking doesn't add value, why not use just the trend filter?

I would venture to guess if you were to apply a realistic slippage number to your monthly rotations, you might even find the ranking/rotation costs you money.

I always say there is nothing wrong with market returns if you can get them on your terms. Contrary to what the buy & hold crowd wants you to believe, there is more performance to be gained by avoiding downside than capturing all the upside.

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Tue Apr 12, 2011 5:04 am

Thank you drm7 - that’s a wonderful effort. In my view it is hugely important on a Forum like this to back up one’s words with actual code, ideas and results of back tests – it is what makes the Forum an interesting place to be. Congratulations.

I intend to revisit your excellent workings in detail over the next few days if I have the time and give it some more detailed and reasoned thought but in the meantime I would add a few comments and thoughts.

1. Data is a problem – always. I seem to recall that the available history for sector SPDRs is fairly limited? But I would again point out the obvious – you need to look back over a far longer history than 10 years. Using whatever data on whatever stock indices you can get. It may or may not change your conclusions but I fear it must be done somehow or another. By way of example when running tests on futures going back to 1970 (on a not dissimilar sort of system) there is a period when it paid handsomely to trade short as well as long; while for the majority of the period shorting produced sparse profits if any. The information was worth knowing. You need to know whether a system has always worked, or not; always worked well –or not. I believe that if a system tests out reasonably well over very long periods one may reasonably draw some sort of tentative conclusions as to its predicted future validity.
2. In my testing, ranking does add value. It certainly comes into its own with a large portfolio of futures where you are restricting overall and sector portfolio risk. You can’t take all trades and it pays to take only the top momentum trades in order. However it is still definitely the case that the bulk of profitability comes from the trend filter not the ranking. The ranking provides the cream on the top.
3. Ranking can also be of value if you decide to add bonds and commodities to the portfolio. There will be periods when you find yourself largely in bonds or largely in commodities. Or indeed largely in stocks. Ranking can be of help in tactical, systematic, global asset allocation.
4. Realistic slippage and commission does not destroy the value of monthly rotation. If that is any help to you. Of course slippage is very much greater on some instruments than others and also depends on your dealing skills and negotiated commission rates.
5. The use of judicious and “freeâ€

drm7
Roundtable Fellow
Roundtable Fellow
Posts: 96
Joined: Sun Apr 20, 2003 9:02 pm
Location: Richmond, VA

Post by drm7 » Tue Apr 12, 2011 9:03 am

Thanks for the quick feedback, Anthony and rgd! I think you both have excellent points.

@rgd - You have a great point, and my (limited) test shows that it may be more cost-effective just to run a straight trend-following system on a broad market index rather than mess with sector ETFs and rankings.

The purpose of this test wasn't to try and "prove" any particular belief - it was more for me to improve my excel modeling skills and test a hypothesis bandied about on this excellent forum.

@Anthony - All great points. I also have your book on my wishlist (so much reading to do...)

I admit that the universe for this experiment was very limited. I wanted to do a test that used free (or already paid-for) resources, like excel and Yahoo! Finance data. I picked the sector SPDRs because they had a relatively long history for ETFs and were a "mutually exclusive and (perhaps) collectively exhaustive" way to chop up the U.S. stock market. There are a lot more ETFs out there, and I know of at least one fund that uses ranking effectively (FUNDX) by casting an extremely wide net.

Here is a counterpoint to consider: would restricting yourself to more recent data in the ETF space be MORE realistic, because the advent of large liquid ETFs may have changed the trading behavior of market indices? Many market pundits have argued that the oil market's behavior changed once the oil ETF gained critical mass. Just a thought.

Also, I also have restrictions in the types of systems that I can actually trade, due to the nature of my day job (regulated broker-dealer). My firm restricts "speculative" trading in futures and requires a minimum 30-day holding period for equities/ETFs, with prior approval of all trades. Therefore, I need a longer-term system.

rgd
Roundtable Knight
Roundtable Knight
Posts: 396
Joined: Thu Aug 10, 2006 5:21 pm

Post by rgd » Tue Apr 12, 2011 1:11 pm

Hi Anthony,

Good feedback. It is interesting to hear relative strength is adding value for you.

I do have a question in applying relative strength to furtures markets:

I trade a portfolio of 60 global futures markets. Right now, for example, I am short the long Gilt and long the US 10 yr.

Under a rotation program, would I look for a "better" short to replace the Gilt, and a "better" long to replace the US 10 yr on a monthly basis?

If I am long soybeans, would a rotation strategy always evaluate whether I should be in meal or oil instead?

I am always curious what others are doing. Based on your statements thus far I might have to reevaluate relative strength.

I do understand applying relative strenght to equities, especially where you would get multiple buy signals on a single day.

However, I am not certain it is very scalable, and liquidity providers have told me that managers who's holdings are essentially known to the public, and trade on a fixed calanedar basis, are getting profiled and picked off by our good friends on Wall Street.

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Tue Apr 12, 2011 4:34 pm

There are so many ways you could approach a momentum strategy.

The first key is to establish the ranking and on what exact basis you intend to rank; are you going to rank long and short separately for instance or will you equate short momentum to long and just have one amalgamated ranking category. Will you define momentum in the same way as FUNDX or does some other formula look better.

Note that FUNDX reallocates the portfolio monthly based on month end rankings. If you adopted a similar approach then yes indeed you would re-arrange your portfolio once a month in accordance with month end rankings. If you trade something like a TMA system which trades every day, then perhaps the extent of momentum ranking would be limited to new trades: don’t take a trade if it is not within the top “Xâ€

rgd
Roundtable Knight
Roundtable Knight
Posts: 396
Joined: Thu Aug 10, 2006 5:21 pm

Post by rgd » Tue Apr 12, 2011 5:12 pm

Thanks Anthony,

I appreciate the feedback.

mojojojo
Roundtable Knight
Roundtable Knight
Posts: 131
Joined: Mon Oct 31, 2005 2:07 pm

Re: ETF Momentum System

Post by mojojojo » Wed Apr 13, 2011 1:33 pm

drm7 wrote:I spent a lot of time on here and have "taken" a lot of great info from some very generous people. This is my humble effort to "give" a little bit, even though this isn't a huge breakthrough.

The excel 2004 attachment runs a "ranking" momentum system on the sector SDPRs. It's simple. Take the top x performing sectors from last month, buy, and hold them for the next month. Rinse, repeat.

I also added a trend filter - i.e., only take ETF trades this month if the SPY (S&P 500) is above its 10 month EMA last month.

I also added a "short" filter - i.e., apply the trend filter, but also short the SPY if it's below the 10 month EMA.

The conclusion? The ranking system doesn't really beat buying and holding the SPY, but applying the filters definitely does - the "straight" filter gives the best MAR and the "short" filter gives the best CAGR.

Hopefully, you can decipher my spaghetti-like IF statements and tweak it to your liking.
You should search for a paper by AQR Capital. They do a study of monthly rotation based off of previous 12month return and I believe Book Value. They test them seperately and together. They test it over stocks within an index, stock indexes, commodities, and I think some fixed income. It is a pretty good paper. I have it some where but can't find it at the moment.

Basically, both methods showed positive expectation; but actually worked better when both were applied together.

drm7
Roundtable Fellow
Roundtable Fellow
Posts: 96
Joined: Sun Apr 20, 2003 9:02 pm
Location: Richmond, VA

Re: ETF Momentum System

Post by drm7 » Wed Apr 13, 2011 3:03 pm

mojojojo wrote:You should search for a paper by AQR Capital. They do a study of monthly rotation based off of previous 12month return and I believe Book Value. They test them seperately and together. They test it over stocks within an index, stock indexes, commodities, and I think some fixed income. It is a pretty good paper. I have it some where but can't find it at the moment.

Basically, both methods showed positive expectation; but actually worked better when both were applied together.
I did read that paper - it was excellent. My focus on this limited experiment was to look at an ETF-specific system, in particular a concept that had been discussed indirectly on other threads.

In my individual stock research (which has been negligible due to the order of magnitude higher expenses required to do it correctly), the aformentioned AQR paper as well as Blackstar Fund's "Does Trendfollowing Work on Stocks" paper (by ecritt on the TB forum) have been my guides.

nodoodahs
Roundtable Knight
Roundtable Knight
Posts: 218
Joined: Wed Aug 09, 2006 4:01 pm

Re: ETF Momentum System

Post by nodoodahs » Fri Apr 15, 2011 10:31 am

drm7 wrote:I spent a lot of time on here and have "taken" a lot of great info from some very generous people. This is my humble effort to "give" a little bit, even though this isn't a huge breakthrough.

The excel 2004 attachment runs a "ranking" momentum system on the sector SDPRs. It's simple. Take the top x performing sectors from last month, buy, and hold them for the next month. Rinse, repeat.

I also added a trend filter - i.e., only take ETF trades this month if the SPY (S&P 500) is above its 10 month EMA last month.

I also added a "short" filter - i.e., apply the trend filter, but also short the SPY if it's below the 10 month EMA.

The conclusion? The ranking system doesn't really beat buying and holding the SPY, but applying the filters definitely does - the "straight" filter gives the best MAR and the "short" filter gives the best CAGR.

Hopefully, you can decipher my spaghetti-like IF statements and tweak it to your liking.
Last time I did this with S&P sector spiders, if ranking alone, it worked best by eliminating only the bottom 1-2 sectors.

Holy grail for U.S. market added value via ranking is at the industry level. Look for data on Fidelity mutual funds, I believe they were called "Fidelity Select" or somesuch nonsense name, they have mutual funds targeted to industries with data going back DECADES. Good for tests.

Delay EVERYTHING on your reading list and search the web for a PDF by Cass in the U.K., which uses ranking+trend filters on MSCI country indices. It'll solve your problem. There are about 3 dozen country-specific ETFs available in U.S. markets. Then you'll have plenty of free time to read everything else.

OTOH if you're FORCED to play in U.S. markets, use a site like Keelix and limit your Momo+TF to the top 100 stocks by liquidity and/or market cap and do that.

Or, if you're not in the mood for DIY, just look for Arrow Funds, they use systematic relative strength and cross-asset allocation to do exactly what you want done. It's a buy it and forget it solution that might be perfect for you in your "day job." Then you will have plenty of time to build systems and test them for when you either (1) switch jobs or (2) retire.

mojojojo
Roundtable Knight
Roundtable Knight
Posts: 131
Joined: Mon Oct 31, 2005 2:07 pm

Re: ETF Momentum System

Post by mojojojo » Mon Apr 18, 2011 2:20 pm

drm7 wrote:
mojojojo wrote:You should search for a paper by AQR Capital. They do a study of monthly rotation based off of previous 12month return and I believe Book Value. They test them seperately and together. They test it over stocks within an index, stock indexes, commodities, and I think some fixed income. It is a pretty good paper. I have it some where but can't find it at the moment.

Basically, both methods showed positive expectation; but actually worked better when both were applied together.
I did read that paper - it was excellent. My focus on this limited experiment was to look at an ETF-specific system, in particular a concept that had been discussed indirectly on other threads.

In my individual stock research (which has been negligible due to the order of magnitude higher expenses required to do it correctly), the aformentioned AQR paper as well as Blackstar Fund's "Does Trendfollowing Work on Stocks" paper (by ecritt on the TB forum) have been my guides.
For ETF-specific system, your best bet is to use the actual indexes instead of the ETF. ETFs are just too new to really get a decent data set. I always try to use ETF data, but typically end up getting the index data instead.

doubleR
Full Member
Full Member
Posts: 15
Joined: Thu Jun 12, 2008 3:25 pm
Location: London

Post by doubleR » Tue Apr 19, 2011 2:12 pm

CASS University paper can be found here

http://www.cass.city.ac.uk/__data/asset ... hes....pdf

ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt » Sun May 01, 2011 4:34 am

Great source for data on "sectors" and/or "industry groups":

http://mba.tuck.dartmouth.edu/pages/fac ... brary.html

Also...the Fidelity select funds are an excellent "proxy" for today's ETF's and most of them have total return data going back to 1987. I've done a tremendous amount of simulation using both Fidelity funds and their "corresponding" ETF's; the results have been virtually identical (and surprisingly good!).

http://finance.yahoo.com/q/hp?s=FSENX+Historical+Prices

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Sun May 01, 2011 5:00 am

Awsome!
Thanks

rgd
Roundtable Knight
Roundtable Knight
Posts: 396
Joined: Thu Aug 10, 2006 5:21 pm

Post by rgd » Sun May 01, 2011 9:35 am

I too use mutual funds as proxies for various ETFs. However, I have found the sector data from the French/Fama link you provide to become more and more unreliable the further back in time one goes (prior to mid 1980s). When compared with data from comparable S&P Groups, it sporadically outperforms in a dramatic and unrealistic fashion. If I recall correctly, the correlations vary widely as well. I found myself discarding the French data and using the S&P Groups data. I know the French data was built from the Compustat data base, but it if was built solely on adjusted stocks prices, it too have found such price series to become more and more unreliable as you move farther back in time. Buyer beware, or maybe better, you get what you pay for.

rossb34
Full Member
Full Member
Posts: 12
Joined: Fri Dec 31, 2010 7:04 pm
Location: Middleton, WI

Post by rossb34 » Tue May 31, 2011 11:33 pm

drm7, thank you for your ideas and testing strategies with us on this forum. Also, thank you to everyone else for contributing to this thread. It has certainly sparked some ideas of my own and I have gone back and tested those similar to drm7. My main reason for doing this was to find a better way to manage my 401k offered by my employer. Originally, I had taken the advice of the financial advisor of a bond/equity mix based on my age, risk factor, etc. I was never really comfortable with someone else making those decisions for me... I saw my father lose 30% of his 401k a few years before he hoped he could retire... I knew there was a better way.

If I trade currencies using a mechanical system, I should be able to do the same with mutual funds. Thanks again to drm7 for sharing his spreadsheet, I was struggling with testing a ranking system with excel.

I ran the tests using vanguard index mutual funds (which claim to have the lowest fees and expenses) and again using iShares ETFs (which can be traded commission free through fidelity).

The conclusions I came up with were similar to others:
Adding a trend filter increases return and reduces drawdown
Adding a ranking system increases return
Ranking and/or Trend Filter beats "buy and hope"

Thanks again for the ideas and contributions.

edit: corrected average return calculation per mojo's comments
Attachments
rev_DRAFT iShares ETF Strategy Testing.xls
(2.79 MiB) Downloaded 557 times
rev_DRAFT Vanguard MF Strategy Testing.xls
(4.54 MiB) Downloaded 520 times
Last edited by rossb34 on Wed Jun 01, 2011 3:04 pm, edited 2 times in total.

mojojojo
Roundtable Knight
Roundtable Knight
Posts: 131
Joined: Mon Oct 31, 2005 2:07 pm

Post by mojojojo » Wed Jun 01, 2011 10:50 am

rossb34

very nice work. I do have one question though.

Going through your ETF strategy spreadsheet, I noticed that your "Average All Returns" columns includes your RS calculation along with your X period returns. What is the reasoning behind that?

mojo

rossb34
Full Member
Full Member
Posts: 12
Joined: Fri Dec 31, 2010 7:04 pm
Location: Middleton, WI

Post by rossb34 » Wed Jun 01, 2011 3:12 pm

mojojojo wrote:rossb34

very nice work. I do have one question though.

Going through your ETF strategy spreadsheet, I noticed that your "Average All Returns" columns includes your RS calculation along with your X period returns. What is the reasoning behind that?

mojo
mojo,

Thank you for catching that, I had not intended to included the ema relative strength in the average return calculations. I have made the correction and uploaded revised workbooks.

That got me thinking of different ways to assign rank. Here are a few of my ideas, what else are people using?

Rank = x-month
Rank = average(1-month, 3-month, 6-month, 12-month)
Rank = average(any combination of x-month return)
Rank = ema relative strength + x-month
Rank = ema relative strength + average(any combination of x-month return)
give each x-month return a weight and use the weighted average


Ross

mojojojo
Roundtable Knight
Roundtable Knight
Posts: 131
Joined: Mon Oct 31, 2005 2:07 pm

Post by mojojojo » Thu Jun 02, 2011 11:29 am

Didn't save the original doc, how did it change the performance?

Also, sounds like you not looking for how to assign rank based on your examples. It seems more like what matrix/value is worth ranking.

There are a ton of different things to rank.

Returns
RSI
% above/below moving average

There are variations to the above. Do you look at absolute values or relative values?
Last edited by mojojojo on Tue Jun 07, 2011 10:12 am, edited 1 time in total.

ratio
Roundtable Knight
Roundtable Knight
Posts: 338
Joined: Sun Jan 15, 2006 11:07 pm
Location: Montreal, Canada

Great site for stock info

Post by ratio » Fri Jun 03, 2011 2:50 pm

http://www.finviz.com


Found that recently, very good web site for US Stock info.


and it's free. I did subscribe to the paying version, but you dont get much more.

Post Reply