In my post about 50/200-day moving average crossovers on the US Dollar I wrote something to the effect of “generally speaking, moving average (MA) crossovers have had some predictive power, but when used alone they are really only useful for staying out of protracted drawdowns…traders are much better served using shorter-term, more active strategies”.
In this post, I want to put my assertion to the test and show how MA crossovers have worked trading the S&P 500. Note that when I talk about MA crossovers here, I’m talking about long-term crossovers like 50 and 200-day MAs (because I have an entirely different take on very short-term crossovers). Here I’ll focus on two classics: 10/50-day MA crossovers and 50/200-day MA crossovers.
[logarithmically scaled]
The chart above shows the S&P 500 (blue) and strategies buying the S&P 500 at the close if the 50-day simple moving average crosses over the 200-day SMA (green) or when the 10-day SMA crosses over the 50-day SMA (red), from 1960 to the present.
To be as fair as possible to moving averages, I’ve assumed (a) frictionless trading, and (b) that the trader earned a return on cash when not invested in the stock market equal to the yield on the nearest 13-week Treasury bill.
As the chart shows, both moving average crossover strategies (but especially 50/200-day crossovers) have done a good job at keeping pace with the broader market over the last 50 or so years. And as the stats below show, both strategies have done an even better job at reducing volatility and drawdowns, and increasing risk-adjusted returns (Sharpe Ratio).
But what both strategies don’t do is significantly increase actual returns. This was the central point of my broad conclusion about MA crossover systems – they are fairly good at protecting investors from protracted downturns, but not very good at actually producing returns.
My suggestion is to either couple moving averages crossovers with other shorter-term indicators, or to ignore them all together in favor of more active strategies (a shameless plug).
Two additional notes:
- This conclusion only extends to very long-term moving averages. As the length of the moving averages becomes shorter, they become more and more of contrarian indicators and less and less of trend indicators.
- Whenever I do a report like this I inevitably get the “yeah, but I don’t use 50/200-day SMAs, I use 47/193-day weighted moving averages, blah, blah, blah.” Look, I’ve tested a gazillion combinations and my conclusion holds true across the board. If you have something in particular that you think breaks this mold, let me know and I’ll run it through the paces.
As I wrote in my last post, market “wisdom” like 50/200-day moving average crossovers continues to live on in the investment community because they’re easy to understand and sound logical…but as I hope I’ve shown here, that doesn’t mean they work as advertised.
Just my 2-cents.
[Edit: click for a summary of all posts in this series on trading the Golden Cross]
Happy Trading,
ms
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Trading Strategies | 30 Comments
-
Subscribe to RSS Feed
Subscribe via Email
My email: michael @ marketsci.com (remove spaces)
You debunked nothing. The 50/200 strategy as you tested it thoroughly dominated the buy and hold. Higher return, lower standard deviation, smaller max drawdown, and less than half the average drawdown. I haven’t personally used this indicator, but after looking at your results I may start.
If by “thorougly dominated” you mean an extra 2%/year then yes, you’re right, we didn’t debunk moving average crossovers. But for my money, that’s not even in the ballpark of good enough.
Maybe I should have added a couple more question marks to the end of the title to make my point more clear. The point of the article is not to say MA crossovers are useless. The point is to say that they’re (a) not as effective as many in the investment community believe, and (b) I believe not as effective as a shorter-term more active strategy.
Thanks for the thoughts.
michael
Michael, Mebane Faber from Engineering Targeted returns and RIsk has a simple model with a price crossover through a 10 month ma and it shows good rsik-adjusted returns…Especially for portfolio’s that are subject to withdrawals, ONLY risk-adjusted returns count. If not ma’s what other strategy can you showcase with better back-test record? What would you consder as the next level up in returs?
RE to Kostas
First off, I have uber-respect for Mebane Faber – I haven’t tested his model though (it’s on the to-do list) so I’m not sure about his performance numbers specifically.
As I discussed in my post “Why I Do What I Do” (see top posts on the top menu), I don’t share strategies that I consider completely “good enough” to trade (b/c this is a zero-sum game). I have however shared a lot of nuggets that someone could use to build their own systems both here (again, see “top posts”) and on the old MarketSci site.
This blog is relatively new, and I’m still in the process of moving old research over (so I don’t have tons of content here yet), but off the cuff, a good one to check out is my post “the VIX is Very Predictable”. Now, I know that that post is about predicting the VIX and not the market, but a savvy reader would remember from my other post “the VIX isn’t Magical” that movements in the VIX and the S&P 500 are highly correlated. So…if we can predict the VIX well (which we can), we can also predict the S&P 500 well. I haven’t shown that yet (planned for a future post), but a little number-crunching on your own will demonstrate it.
Is that VIX driven strategy “good enough” in my eyes? Still no (I have high standards), but it’s a mile better than the 50/200-day crossovers I discussed here.
Thanks for the comments.
michael
Agree, crossovers as a system are worthless. However they can serve as great trend indicators that keep you out of prolonged draw downs, just as your post clearly concluded. Good post but I would prefer you not share such information in a zero sum game. Our living is based on 80% of all the investors being dumb. I’m kidding. Actually I’m not kidding :)
Also remember that compounding matters…especially over a span of 50 years. 6.4-6.5% annually and 8.3% annually gives more than the double in return over 50 years (ex taxes etc.)
RE to Rene: Absolutely…I completely agree – that’s the power of compounding. But wrap your imagination around this – what if we could outperform the market by 5, 10, 15, 20%+ annually? Over a span of even 10 or 20 years…wowzahs
Also true :) Thx for at great free site!!
Another thing to keep in mind is that you can short the SPX on the 50/200/SMA downtrends instead of going to treasuries. That strategy does not work as well back to 1960 (1978-1999 was horrible), but it worked great after 1999, because you are making nice returns during most of the .com bust (33% profit from 10/00 – 5/03) and also over the last few months (37% profit from 12/07 – today).
Also important to note is that buy and hold is more tax-efficient, so the SMA system’s return advantage is lessened when used in a taxable account. But lower volatility is also important to keep investors happy and SMA definitely wins in that respect (especially if you expect more long downturns in the years ahead).
RE to Erik: agreed on all accounts. Smart comments…I appreciate it.
michael
I am not a long-term trader myself, but I must say that in the long-term 50/200 crossover works really well. I mean, it’s important to be upfront as to what it can do for you. And what it does best is tell you when the market is starting to trend up (for long position entry) and when the market stops trending up (for the long position exit). So, if you cared to watch this simple indicator on a regular basis, you would have got a clear signal to get out on Jan 7, 2008 – and that would have saved you an agony of watching the mkts half in value in the next 10 months.
The problem with many traders today, even the long-term investors, is they tend to ignore these simple and shall I say “primitive” indicators in favor of their sophisticated tools and trading techniques (me included!). But my conclusion gents is: Keep It Simple. If those high-flying hedge-fund mavericks considered this simple indicator, they would have exited their long equity positions in Jan and saved their investors billions of dollars. But no, they thought they knew better.
The irony is, as long as majority of traders ignore this very powerful tool, and they likely will in favor of some new fancy indicator, 50/200 indicator will continue to give very good signals.
Happy trading!
RE to RK: I agree with you to a point. Yes, long-term MA crossovers are a world better than what the vast majority of self-directed traders (and even financial professionals) are using. Enough studies have been done showing average trader performance to prove that. Point well taken.
But I also think that a solid short-term signal wins the day (based on my own performance and a handful of other folks whose audited results consistently outperform the broader markets).
I think it’s dangerous to use the fact that this indicator would have protected investors from the October crash as justification for its use. Maybe a lunar cycle indicator would have done the same, but that doesn’t mean I’m going to use it in my own portfolio. The reason that I do long-term historical studies is to move the collective discussion away from “it worked this time” towards “how has it worked every time”.
Having said all that, again, I completely agree that this very simple indicator is a ton better than 99% of what’s out there.
Thanks for the comment!
michael
What’s really surprising is when you have a “buy” signal from either
of the 2 crossover methodolgies, and buy the newly created
ultra long ETF for the SPY (SDS) the results are quite amazing.
“…but when used alone they are really only useful for staying out of protracted drawdowns”
I always shake my head when I hear something like this because of how much wealth could be saved if people could just avoid these “protracted drawdowns”, especially later in life when they have the most to lose.
Any investing strategy should not just take a set amount and backtest but should be based upon real-world investing. This means that it should include the fact that not only do investors have more to lose later in their investing lives but that they can’t afford to take such a loss, as we are seeing now.
A simple investing method that takes an investor out of the market when it drops below the 200 day moving average would not only generate good profits but would save a tremendous amount of wealth in bad markets.
RE to Fred Voetsch: I’m not saying that long-term trend following systems like this are without use. In fact, 50/200-day crossovers are included in our own State of the Market report (http://marketsci.wordpress.com/state-of-the-market/). What I am saying is that relative to (a) some extremely simple short-term systems we’ve talked about on this blog, and (b) our own programs that we offer through managed accounts, I think investors could do significantly better.
That includes during bear markets. For example, during this current bear market, despite being one of the worse in the history of the markets, we’ve able to make significant gains…in many ways it’s been a boon for us. These long-term trend followers are great defensive systems, but they’ll never be able to accomplish that.
michael
Dear Michael Stokes,
Let me say, only you and very few others actually have websites worth referring to on a daily basis.
I have a few questions for you have a moment:
1. Is 50, 200 the most robust parameter you have seen across all asset classes?
2. Would you care to share other useful MA parameters? Do you see any advantages to using 3 MA’s in your testing?
3. Which software and data provider do you use to conduct these kinds of tests? Could you suggest any offline or online packages for novice users who want to conduct this kind of testing?
Thank you very much..
RE to SS: thanks for the kind words.
RE: “the most robust parameter” – no, but it’s a generally accepted pair in the investment community so I usually use it to represent a long-term trend following strategy.
RE: “other useful MA parameters” – this is a complicated question – different MA pairs are measuring very different things – as MA’s become shorter, they tend to become more and more contrarian (mean-reversion) and less trend following. We’ve talked about this previously, for example:
http://marketsci.wordpress.com/2009/01/13/the-moving-average-spectrum-part-i/
http://marketsci.wordpress.com/2009/01/18/the-moving-average-spectrum-%e2%80%93-part-ii/
RE: “software and data” – talked about this previously here as well:
http://marketsci.wordpress.com/2009/01/06/faq-software-and-data/
Hope that helps!
michael
This is interesting, because I’ve run the same test with the same set of data (S&P500 daily/YHOO provider adj. close as time series) and found that buy and hold beat 50/200 crossover over the length of its history. Any way you could share the excel file for comparison?
P.S. Are you also shorting on crosses in the opposite direction?
Or just buying on cross in one direction, and selling on opposite cross?
Thanks,
Pat
RE to Pat: long-only. Two questions: (a) SMA not EMA right?, and (b) did you account for the return on cash? (there’s a paragraph somewhere in there that talks about that). If you’re still having trouble, let me know and I’ll try to get something together for you. michael
Hi and thanks for the fast reply.
I am referring to SMA only. In fact, I have just rerun the data using ^GSPC
from 01/03/1950 to 12/21/2007 which coincides with the last sell date on the MA Long Only
crossover system. I could have used more recent data, but doubt it would change much.
I do not use short crossing data, as it actually performed worse over the long run, thereby it would bring down the overall results if used. As it stands, using the data I described,
the long only crossover produced a 5,557% gain, while the buy&hold over the same period, with
coinciding end date as last sell date of long only system netting a 8,810% gain. Clearly
trouncing the crossover system, unless I am missing something. That’s why I was hoping to
compare at least spreadsheet information about the individual dates and corresponding returns
for the 50/200 crossover system.
Admittedly, I do not take into account any returns on cash during the neutral periods where you would be out of the market. Although, it’s hard to take that into account objectively, as cash or tbill rate of return would vary of those periods, depeding on what was used as basis.
Do the numbers not including cash return, generally agree with yours?
I’ve done other types of simulations along the same lines, whereby you run random sets of returns over some period (like 5 yrs) and compare to buy & hold then generate a matrix or surface plot of returns. Buy and hold generally trounces under that scenario as well.
P.S. It’s also not really fair to account for return on cash during holding periods, since not only is it somewhat subjective, but in fairness, you would also have to account for the tax structure on the crossover system, since not all trades fall under long term tax rates.
RE to Pat: I’m not sure to what degree the results were influenced by return on cash. I’ll have to rerun the numbers and take a look see. Give me a couple of days and I’ll email you something.
I do have to however humbly disagree with you that it’s not fair/important to account for cash and time spent out of the market when assessing a strategy – especially one of a long-term nature like this one. I think it’s important for a number of reasons: (a) reduced risk of inherently unpredictable market events, (b) investment capital free for other opportunities, and/or (c) the more obvious one – the actual return it generates. I think not accounting for return on cash paints an inaccurate picture of the worth of a strategy.
I don’t think that’s the same thing as accounting for tax consequences. Return on cash is generalizable to all investors. Taxes are important obviously when assessing any approach to the markets, but they are very specific to the specific trader.
Having said that, reasonable people could disagree on the best approach to accounting for that return on cash (I use half the 13-week treasury at that moment in history).
michael
It’s an interesting and debatable point.
However, you raise another interesting peripheral point, which is–
are trading simulators subjective and inaccurate?
Because I cannot think of any trading simulator that accounts
for return on cash during holding periods.
Most traders I would suspect do not park their cash in t-bills during transitions
(it would be nice if brokers had such a feature), they generally let it sit in
money market funds, which are generally not all that great.
So, in the grand scheme of things, I think both arguments about
how tax as well as cash returns affect the results of
backtesting comparisons vs. buy and hold conclusions; introduce some
degree of uncertainty and subjectiveness in the results.
That is why, philosophically, it makes more objective sense to
compare the results directly (without including tax or cash return implications).
Although, I see your points, and they are well taken.
Anyways, it is refreshing to see one post who actually does some actual quantitative work to back up their assertions. I look forward to seeing the data
on returns sans-cash neutral periods, and also more of your insights!
Pat
Try back testing swing trading using a 7/21 SMA and Slow Stochastics for long short positions in the SP500 using double long and short ETFs. Go long when the 7 MA>21MA and K>25 and go short when 7MA<21MA and K has traced above 75 then back down again to 75. Buy the short instrument on the downslope of K at 75. Buy the long instrument on the uptrace of K as it passes below 25 then up to 25. Set tight stops in case something goes wrong and just roll with the cash flowing in. This works when the VIX is high, say above 35. I made a lot in the early winter this way with pretty small starting positions that I kept compounding with the trades. With a 3500 starting position it was not uncommon to make $1800 profit in two weeks with about 10 trades. If you don’t believe me backtest it or try it right now.
Interesting analysis.. I also did some research that I put together on the 200 day moving average and its significance along with our current market condition in that we are approaching the 200 day moving average. You might be interested in it.
http://www.nakedhedgefund.com/finance/hello-200-day-moving-average-long-time-no-see-some-analysis/
If anyone is still reading … do these results include dividends, and if not, how much does it help BH when you include them?
RE to danbek: yes, both B&H and the strategis already include the impact of dividends. michael
Hi,
Has the 50 day now crossed over the 200 around SP 1000?
Out at 1400+ and back in at 1000 without the heartburn? But I have my doubts about staying power of this market this may be held up by one bailout/rebate after another.