It works!Posts by Ted, MJG, and Flack on the 200-day Simple Moving Average (SMA) technical signal got me curious:
Most recently:
Hulbert's Test of the 200 Day Technical Signal200 Day Versus 10 Month Moving AveragesAnd then from 2011:
The Effectiveness of Simple Moving Averages
We’ll use a 200-day Simple Moving Average on the SPY (S&P 500 Index Exchange Traded Fund).
We buy at the end of the month when the price has closed above the moving average.
We sell at the end of the month when the price has closed below the moving average.
(Does it get any easier than this?)
Only because the monthly database was readily available, I used the 10 mo SMA, but in the same manner Flack describes. If current month was above 10 mo SMA, the model was all-in SPY the next month. If it was below, the model was all-in 30 day TBill.
Here are the tell-tale graphics on why Flack's approach works:
The approach
catches most of the upside and
avoids most of the downside.
Here's a look at lifetime performance, all measured from 10 mo after SPY inception, about 20 years ago:
Significantly higher absolute growth (by more than 200%) and APR (by more than 2%) than either SPY buy-and-hold or a 60/40 SPY/TBill fixed portfolio.
Significantly lower volatility than SPY buy-and-hold. In fact, the downside standard deviation DSDEV is almost half! Making risk adjusted returns twice a buy-and-hold strategy...both Sharpe and Sortino.
Ulcer Index is substantially lower than either SPY or 60/40 portfolios. And, perhaps best of all, the maximum draw down is only -17% versus -51% and -34%, respectively.
And, a look at rolling 3 and 5 year periods:
In 170 rolling 5-year periods, the approach beat the fixed 60/40 returns 100% of the time and never drew a loss. It beat 83% of 194 3-year periods as well...and it never lost more than 1%.
So, unless he is talking about relatively short periods, I do not understand Mr. Hulbert's disappointment, as summarized by MJG:
Hulbert is not happy with the results. Over the entire history of his data set, he finds that the indicator does yeomen work, but it has failed over more recent selected timeframes.
How can anyone be disappointed with performance of this method? Results of the SMA dynamic allocation method are remarkable. It produces healthy returns with much less volatility.
So, hats off to Flack. My belief continues to be that investors owe it to themselves to
actively manage their portfolios against downside risk.
MJG, I can feel you cringing now, but I will keep an open mind and be sensitive to opposing views
.
Ted thanks again for link that led me to Faber's paper
A Quantitative Approach to Tactical Asset Allocation.
Comments
There's a number of these timing funds, but I haven't seen one that has pulled off the strategy consistently. I'm very open to alternate strategies, but I haven't found one of these timing funds (such as the Toews funds) that I'm interested in. I think people can certainly pull off these strategies, but I'm not that interested in the funds yet.
Thanks for running the stats on the 10-month simple moving average.
Actually, I don’t know how to generate these numbers (I’ve never learned
how to do a simple spreadsheet). I appreciate you taking the time to do this.
As you may know, I do an adult education investing class a few times a year.
Some people have asked me if they can apply a 200-day or 10-month
simple moving average to their funds or individual stocks to generate
buy/sell signals.
I start by telling them no. Then I explain (which I thought everyone realized)
that every stock and fund has its own ‘personality’, its own price/time rhythm and
that it’s necessary to find that rhythm and the corresponding moving averages
that best match it.
In my classes, I show people how they can easily reduce their market risk and
generate decent returns by being proactive. I suggest that they can do this with
ETFs – and the fewer, the better. Take for example using just ONE of these
for stock market exposure – VTI, SPY, VYM or SDY. Each of these so closely
follows the S&P 500 that they generate almost identical buy/sell signals
as the index.
Now some people want foreign exposure and there are times when they could
add an ETF to satisfy this craving. But I always ask myself if this is necessary.
I don’t believe it is, if the objective is to keep investing as simple as possible
while generating market-beating returns with reduced overall risk.
From all appearances, a great number of investors are hooked on the hoax
that they MUST diversify by owning multiple stocks and/or funds. They fell for
this hoax early on and they can’t shake it; they can’t step away and with fresh
thinking ask, “How can I construct the simplest portfolio and strategy?”
By the way, Hulbert’s study used the Dow (which I would not recommend)
and generated a buy/sell signal the DAY that the price crossed the moving average.
That created too many whipsaws. You did it the best way, which is to use the
end of the month – although I have to tell you that there is nothing magical
about 30 or 31 days, so personally I allow an extra day or even two if the price is
hovering at the moving average.
Thanks again for your response and the work you did.
Flack
Hey, I looked at a variation of the portfolio above using iShares Barclays 7-10 Year Treasury Bond IEF instead of cash. And the results are even better!
The window is a little shorter, about 10 years since IEF inception. Here is the SMA timing chart for this period, again using SPY like you recommend, along with performance plot:
The attendant lifetime performance:
And, rolling 3- and 5-year period performance:
I realize bonds have been up during this period, but still, I'm impressed with the results and the simple implementation of the SMA timing method you champion. Thanks!
http://stockcharts.com/h-sc/ui?s=IWM&p=D&b=5&g=0&id=p21876688241
http://stockcharts.com/h-sc/ui?s=SPY&p=D&b=5&g=0&id=p06780032430
Sorry for my late entry into this discussion. I was away visiting a couple of baseball spring training facilities near Phoenix. Great fun; a delightful regression into the past. It’s a renewal thing for me.
I do have a question for the group that I will pose at the end of my post.
As you know, I am not a Technical analyses fan. I heavily discount (but not eliminate) most technical analysis when making investment decisions. My bottom-line assessment is that almost all the endless list of candidate methods are bogus. They have little fundamental merit, yield numerous false trading signals, and have a nonexistent out-of-sample experimental database.
Although I don’t trust most technical analyses formulas or technical analysts and their writings, I do acknowledge that exceptions do exist. Two exceptions come immediately to mind.
First, I have always liked and endorsed Tom Bulkowski’s technical books and his informative website. His “Encyclopedia of Chart Patterns” is a classic. The book is comprehensive and somewhat overwhelming in its data sets. It includes summary tables that rank charting effectiveness and pattern performance statistics. It is indeed encyclopedic by design and in execution.
Bulkowski’s website is just as comprehensive as his many books. It is a treasure chest for Technical analyses devotees. Please visit it. Here is a Link to that outstanding resource:
http://thepatternsite.com/
Second, I have consistently recognized that Simple Moving Average (SMA) methods do offer some market momentum guidance. Although I have never advocated using SMAs for definitive decision making during a hard crossover point, I do deploy a form of SMA in a composite forecasting criteria soft sense. I do not accept a single market signal for investment decision making purposes; the matter is far too complex, so I deploy a multifactor model.
I outlined a provisional multifactor model in a MFO posting in June, 2011. I still believe it captures much of my thinking on the matter. Here is the Link to that internal submittal:
http://www.mutualfundobserver.com/discuss/index.php?p=/discussion/634/an-informal-recessionmarket-trend-model/p1
I always considered the model to be a provisional tentative attempt to forecast market movement; it is subject to modification dependent upon accumulating data and dynamic market changes. It is experimental in nature.
The fundamental goal was to identify and use a minimum number of econometric and market indicators to signal broad market direction. The generic six factors were: (1) Fiscal, (2) Valuation, (3) Momentum, (4) Macroeconomic, (5) Liquidity, and (6) Sentiment (Psychology). A single metric was selected to measure the state of each factor with respect to the directional likelihood of the equity marketplace.
The referenced MFO submittal identifies the chosen metrics. The models validity is still an unsettled issue; it remains an experimental project.
The momentum component uses a variant of the SMA concept. I lifted its description from the earlier MFO submittal as follows:
“The Momentum (3) factor is evaluated by examining the difference between the 65-day and the 200-day S&P 500 Index moving average plots. The momentum metric is a primary factor and controls, along with the Fiscal factor, if the signals from the other four factors will be implemented. If the 65-day moving average penetrates the 200-day data series on the downside, a sell signal is generated. The reverse is actionable on a surging equity market. A 15 % portfolio realignment is executed if this test is satisfied. “
I selected this particular variant mostly because I am lazy and continually reduce my investment time commitments. These data are conveniently reported in the Monday edition of the Wall Street Journal, so the data are easily available.
Understand that I am not a slave to this methodology; it is highly experimental. I interpret the outcomes in a soft sense. I reserve the option to override the model numbers with personal judgment sharpened and shaped by experience. Gut instincts matter.
I hope this clarifies my position on SMA procedures.
My question to you all is: “ Does anyone have insights or experience with the differential form of the SMAs that I elected to use?”.
I selected that particular formulation anticipating that it would be more sensitive (quicker to react) than a single SMA representation. I have not tested that assertion.
Best Wishes.
Thanks for joining the discussion.
From your most recent post –
”…tentative attempt to forecast market movement.”
“…composite forecasting criteria…”
“…to forecast market movement…”
I find it interesting that you attempt to forecast the market,
even though you have posted numerous times that forecasting
is an unproductive activity.
To keep this short, the single Simple Moving Average device
that we’ve been discussing is NOT a forecasting device.
It identifies and follows the major market moves.
Every financial advisor on this planet recognizes that when the market (S&P)
is above the 200-day simple moving average we’re in a bull market and when it’s
below, we’re in a bear market.
BTW, I couldn’t locate the Monday WSJ data you mentioned.
Flack
Hi Flack,
You made too cursory an examination of the WSJ. You overlooked the graphs.
The Moving Average data that I referenced in my post is exactly where it always is in the Journal’s Monday edition. It is placed in the Market Data section of the paper. In the Monday, February 25, 2013 issue the plotted data is displayed on page C4. The two figures at the top of that page show the referenced Moving Averages for both the Dow Jones Industrials and the S&P 500 Index. I use this material for several of my needed inputs.
I still do believe that forecasting with any accuracy is a daunting challenge, perhaps ultimately an impossible task. I noted that my modeling in that arena is experimental; to date, my efforts have not been entirely successful so I do not fully trust its current outputs. Reliability remains an issue. But I’m a good soldier, so I will continue the march.
I never intended the forecast to function as a market timing tool. It’ll never be precise enough to serve that purpose. I typically rebalance my portfolio about once annually. My goal was to project equity returns using this experimental model and compare it with two other simpler models that I also evaluate. Based on these soft projections, I adjust my asset allocation just a little to sometimes reflect their composite outputs in the portfolio rebalancing effort.
Note that I said “sometimes”. To refresh your memory, I said in my original posting that “I reserve the option to override the model numbers with personal judgment sharpened and shaped by experience. Gut instincts matter.”
I have never completely trusted forecasts so I have always made rather modest, incremental changes to my asset allocation distribution.
Congratulations on your successful application of the S&P Moving Average. I wish you continued success. Stay alert, things change suddenly.
Best Regards.
Thank you for the additional explanation of your investing methodology.
With your help, I found the WSJ charting data.
It looks much like Yahoo's charts. I prefer Stockcharts.com for their cleaner graphics,
although its data only goes back 3 years.
Barchart.com goes back some 25 years but it's clunky - that's a technical term
"Stay alert, things change suddenly."
Indeed they do, which is why I still do some trading.
My best,
Flack
The timing curve is the same as above, but the result is quite different, as seen in the portfolio growth plot below and corresponding performance table:
Flack's dynamic timing method still trumps either the SPY or stock/bond 60/40 buy-and-hold portfolios, but this time with much higher volatility, greater max drawdown, and with somewhat lower overall growth. This result highlights the potential vulnerability of the method, the so-called "head-fakes" and getting out of sync with the market. In the long-only portfolio, having a string of back-to-back reversals (see crossover symbols on timing charts above, particularly Jul-Oct '10) means you may not capture a near-term gain, but if you are short...you can get whiplash.
Here are links to couple informative presentations by Mebane Faber: Dynamic Risk Parity Part 1 and Dynamic Risk Parity Part II.
It's basically about how to survive down-turns (bubbles) to invest another day, or like Matthew Moran of ASTON/River Road Long-Short ARLSX explained during the recent MFO teleconference: "Stay out of the way of freight trains!"
In any case, Flack, thanks again. MJG, thanks for link to Bulkowski's Blog. Scott, I will look further at the shorter SMA period. I think the ARLSX folks swear by the 50 day SMA of the Russell 3000. Hank, what can I say, I guess I'm a quant at heart.
Even if the experience is a bit different, here is look at using the inverse fund SH. Unfortunately, it has only been around since 2006. Timing certainly paid-off in 2007, but again experienced whiplash in 2010. Here are the graphics:
And corresponding performance:
While beat pure SPY buy-and-hold, still had large drawdown compared to long only approach.
The scheme is known to reduce the downside risk but I am afraid you conclusions is overly clouded by 2008. If we do not experience such devastating black-swan like event soon, the strategy is likely to lag. I recommend you run your analysis on a much longer time. Preferably, I would look at the rolling 5 or 10 year results. In other words, there is more than one way to skin this cat. You only skinned one way.