200-day moving average market timing

[JW-Retired]

Special note from Boglehead Administrators: Bogleheads are not market-timers. Mr. Bogle's mantra is "stay-the-course."

I use the 200 day moving average, as crossed over by the net asset value.

R48,
I never paid much attention to tech analysis maybe because of Louis Rukeyser's disparaging attitude toward "Elves" on Wall Street Week in the 1970's. However, I roughly tried your 200day/NAV crossing system out on VEIEX with startling good results. Got interested so I've done some more careful numerical simulation programming of the system on a number of funds. Used daily closing prices from Yahoo and made the basic assumption that if the closing NAV passed through the 200-day average, with a margin of 1%, I would trade. Looked at 7 ETFs/funds. VEIEX= emerg mkts, SPY= sp500 ETF, VEXMX= extended mkt, QQQQ = NASDAQ-100 ETF, DIA= DOW 30 ETF, XLE= energy ETF, XLF= financial ETF.

Most of the Yahoo ETF data only goes back to about 1998. I used a start date at the first crossover buy opportunity in the data (varied from 1992 to 2002) and invested $10000 at that time. Buys/sells then happened automatically in the program. To be specific, if the day closing price met the criteria, then I would trade at the next day's closing price. You could more or less implement this system just by checking the closing prices every evening. Results are as follows:

Fund gotin-date price-in shares invested gotoutdate cur.value .. buy&holdvalue #of-trades
VEIEX 1996.58 .. 11.50 . 869.57 . 10000 . 2008.4326 . 40767.72 .... 13530.5 ... 22
SPY . 1995.009 . 46.00 . 217.39 . 10000 . 2007.9535 . 27289.84 .... 20105.3 ... 42
VEXMX 1992.58 .. 16.00 . 625.00 . 10000 . 2008.4791 . 17287.21 .... 15057.0 ... 66
QQQQ. 2001.93 . 42.91 . 233.05 . 10000 . 2008.6626 . 11071.13 ..... 7170.9 ... 28
DIA . 1998.962 . 88.72 . 112.71 . 10000 . 2008.0055 .. 7534.46 .... 10023.0 ... 56
XLE . 1999.781 . 26.81 . 373.00 . 10000 . 2008.5585 . 18401.06 .... 18657.5 ... 42
XLF . 2000.222 . 23.77 . 420.70 . 10000 . 2007.5236 .. 9547.64 ..... 5805.7.... 38
...................................................................................._________......_______.......
......................................................................totals.... $131,899 ..... $90,350 .....

Above table shows the buy-in dates, price, and #shares you got for $10K, along with the current value you would have if (1) you used the system, and (2) you bought and held until today. The system has you all in cash right now, having got you out somewhere between mid-2007 & summer 2008, depending on the fund, and keeping you out. Previously it had you doing an average of around 3-6 trades per year. The trades come in clusters.

Overall results are pretty good compared to buy-and-hold. In terms of the totals of the 7 funds, you are 46% better off to have used this system compared to buy-and-hold. In 5 of the 7 funds have you are much better off using the system, one (XLE) is about even, and only in DIA are you worse off to a great degree. Looking at the performance over time the system loses a lot of DIA money in 2000-2002 because the DIA price is basically stagnant but jittery, so you there are a lot of trades where you just keep losing the 2% tolerance band every time you get in and out. It adds up to a lot. Maybe there is some way you could avoid that.

Anyway, performance is not too bad at all given the market. On the other hand, any system that got you out of the market the last few of months is going to look good. Looking at the values over time the buy&hold would be better up until the last few months of the bear market.

This analysis ignores CG, dividends, and taxes.

Bottom line is I'm thinking maybe this sort of market timing is not such a bad idea. It does protect you from times such as now. Anybody have some thoughts on this?
JW

[MWCA]

There is no fool proof system for market timing. If there was someone would have figured it out by now and everyone would use it.

[Avo]

Here is an interesting article on market timing from Merriman, a firm that's been using timing for 25 years (though they've recently de-emphasized this approach):

http://www.fundadvice.com/tools/hotline ... ned-2.html

[dmcmahon]

Try back-testing it against the S&P 500 and make sure to include the full 40 years starting from, say, 1968. Timing looks good when there are relatively long multi-year trends to follow - it looks less good when there's a lot of choppy up-and-down actin.

[james22]

We need to simply remember that downside violations of 200 month moving averages are quite the rare beasts. As we suggested when we last covered this topic, the only times we’ve seen occurrences where this has happened have been accompanied by periods of very meaningful fundamental economic challenge. As you can see below, we’re there again.

So, either the forced liquidation and deleveraging of the moment is anomalistic enough that it is not allowing the financial markets to more properly discount forward economic reality to come in current prices, or this very deleveraging process itself will further contribute to negative real world economic outcomes ahead and the markets are suggesting as much.

http://contraryinvestor.com/mo.htm

[Ariel]

We need to simply remember that downside violations of 200 month moving averages are quite the rare beasts. As we suggested when we last covered this topic, the only times we’ve seen occurrences where this has happened have been accompanied by periods of very meaningful fundamental economic challenge. As you can see below, we’re there again.

So, either the forced liquidation and deleveraging of the moment is anomalistic enough that it is not allowing the financial markets to more properly discount forward economic reality to come in current prices, or this very deleveraging process itself will further contribute to negative real world economic outcomes ahead and the markets are suggesting as much.

http://contraryinvestor.com/mo.htm

Never saw a 200-month moving average before. Interesting, and troubling

[Erwin]

JW: I have been looking at this same subject for awhile, doing nothing, but wondering if it could fit my situation, now that I am retired and cannot affort to wait 10-20 years for the market to recover. In the mean time I have seen my equity portfolio drop by 30-40%.
I suggest that you read "A Quantitative Approach to Tactical Asset Allocation" by MEBANE T. FABER. Yo can search it in the Internet. Also J. Segal in his book "Stocks for the Long Run" dedicates a chapter to this subject.
My conclusion from reading these articles plus others, is that you certainly reduce risk, i.e., standard deviation, but potentially at a price of reducing return. The system works great during times of sharp price decline (like now) by it is not that effective during times when prices zig zag on the way up, when the system forces you to exist the market at a price typically lower than the one you enter back. For this reason, if you are investing for the long run, depending on which period of time you pick (as J. Segal shows in his book), timing the market with this tool may do from about the same than buy and hold, to even hurt you.
My own observations is that if you had used the 200 d MA to time the market during this crisis, you would be out of the stock market sometime late 2007 and out of commodities early summer 2008, missing most of the loses. You can try it by yourself by ploting any of the indexes in financeyahoo. com
I have seen data from an investor in another forum, see QPP in Yahoo's forums, who back tested a portfolio consisting of Vanguard Index Funds (60% stocks and 40% bonds) from July 1999 through August 2008, and got the following results (see first returns, followed by standard deviations): B&H: 10.6%, 9.3%; Rebalanced annually 9.6%, 11.7%; 200 d MA: 9.1%, 7.6%.
So you decide!
Erwin

[Hemispheres]

http://contraryinvestor.com/mo.htm

Why should anyone take seriously an article that uses a linear y-axis on a long term stock chart?

[Erwin]

Please correct in previous text: Rabalanced annually: 9.6% return, 8.1% standard deviation. Erwin

[james22]

http://contraryinvestor.com/mo.htm

Why should anyone take seriously an article that uses a linear y-axis on a long term stock chart?

Contrary Investor is written, edited and published by a very small group of "real world" institutional buy-side portfolio managers and analysts with, at minimum, 20 years of individual Street experience. Our credentials include CFA, CPA and CFP, as well as the obligatory MBA's in Finance. We are all either partners or employees of institutions with at least $1 billion under management.

Contrary Investor is our vehicle for providing what we believe is institutional quality financial market research, analysis and commentary that is characterized by honesty, integrity and credibility. We live the business and hope to bring what we learn from our daily experiences to our work at Contrary Investor. Having checked our egos at the door many moons ago, we hope to allow our work to stand on its own and speak for itself, without the personal promotion of any one individual getting in the way. No investment guru's here. Just lifelong students of and participants in an ever-changing financial marketplace.

[DP]

Hi,

Try back-testing it against the S&P 500 and make sure to include the full 40 years starting from, say, 1968. Timing looks good when there are relatively long multi-year trends to follow - it looks less good when there's a lot of choppy up-and-down actin.

In the paper by Mebane Faber, already referenced, the 200 moving average was tested on a monthly basis against the SP500, or a proxy for it going back 100 years. For good measure this was considered in sample, so it was then tested on 20 other markets, going back from 36 to 100 years. In all but 1 case risk was reduced and in most cases returns were higher. Is that long enough?

I tested a number of moving averages with a diversified portfolio over the past 11 years. In ALL cases the moving averages beat buy and hold on a total and risk adjusted basis. This was a surprise, I really thought that there would be at least some moving averages that underperformed.

Such methods are not perfect, nothing is. From 2003 through 2007 buy and hold outperformed the best moving averages that I tested, but not by much. But during the bear markets it isn't close. Moving averages are just a tool that can be used to reduce risk, and it's almost certain that it will not be successful at this over all time frames ... but then I would say buy and hold is not successful over all time frames either.

Don

[gummy]

Intrigued by the notion that a 200-day MA is better than Buy&Hold, I done did this:
http://www.gummy-stuff.org/200-day-MA.htm

I find that over the last 10 years, the MA portfolio done good (but "the market" didn't).
Aaah, but over the previous 10 years, Buy&Hold portfolio done good (and so did the "the market").

Fascinating, eh?

[TheEternalVortex]

Why are you using 200 days? Maybe 150 days is better? Or 225 days? Or 134 days?

Maybe you should choose the value that would have been optimal, given available data...

[daryll40]

I started investing by using "Dick Fabian's Telephone Switch Newsletter" in the late 1980s. Later his son, Doug, took over and Dick retired and was seen, literally, huckstering some "learn to play the piano by ear" thing on late night TV. Order before midnight tonight!

The point of the newsletter was using a 39 week moving average, almost the same as the 200 day moving average. It looked good on paper and would "save" the investor from terrible bear markets.

Unfortunately in the real world there were terrible whipsaws. This was before ETFs so you got end of day mutual fund pricing which made matters worse. So instead of death by the guillotine of a bear market, it was death by a thousand cuts of whipsaws and constantly being on the wrong side of the market (out when markets rose and in when markets started going down).

This was my first inkling that maybe you had to hold on for the best long term returns. I did not know the term "market timing", but I learned that it didn't work.

[gummy]

Why are you using 200 days? Maybe 150 days is better? Or 225 days? Or 134 days?

Maybe you should choose the value that would have been optimal, given available data...

Did you take a peek at the link? You can stick in 150, 225, 134 ... or 2.

[DP]

Gummy,
Great spreadsheet. I think you show that this does not work well for stocks (and DJIA). Have you tried moving average crossovers against indices? I tried exponential moving averages from 1 to 200 vs. averages of 10 to 300 against a diversified portfolio: VTSMX, VGTSX, ^DJC, VGSIX, VISVX, VEIEX over the time frame from May 1998 to Nov 2008. In all cases I found moving averages to beat buy and hold.

I'm surprised the results should be so different, but moving average crossovers trade less frequently and perhaps most indices tend to trend more than stocks, and the DJIA.

Don

[Erwin]

So you held until now and have absorbed 30-40% loss on your ETF portfolio. Now you "see the light" and decide to switch your strategy to using the 200 d MA to determine exit and entry points. But all your ETFs are way below the 200 d MA, so what do you do? You sell everything and wait until the 200 d MA is above the current price to buy back or you decide that it is too late to switch and ride the market? Erwin

[gummy]

Gummy,
Great spreadsheet. I think you show that this does not work well for stocks (and DJIA). Have you tried moving average crossovers against indices? I tried exponential moving averages from 1 to 200 vs. averages of 10 to 300 against a diversified portfolio: VTSMX, VGTSX, ^DJC, VGSIX, VISVX, VEIEX over the time frame from May 1998 to Nov 2008. In all cases I found moving averages to beat buy and hold.

I'm surprised the results should be so different, but moving average crossovers trade less frequently and perhaps most indices tend to trend more than stocks, and the DJIA.

Don

The plot thickens!
I done some Vanguard stuff and get this:
http://www.gummy-stuff.org/200-day-MA.htm#DP
For example:

Mamma mia! Quite a difference, eh?

[rwwoods]

Take the following for what its worth. I am only reporting what was published. Colby & Meyers published in The Encyclopedia of Technical Market Indicators (1988) a study of several market indicators. One of the indicators is a simple moving average. They varied the number of weeks to find the time that provided the best profit of trading the NYSE index over 19 years from 1968 through 1986. They traded both long and short depending upon the signal. A buy-and-hold strategy provide a simple average annual return of 8.31%. Using a 45 week simple moving average provided a 10.93% return or an excess of 2.62% per yr. There were 50 trades, but trading costs were not included. If one overweighted or underweighted a NYSE index holding by X%, then the excess return would have been 2.62/X% per year. Thus a 10% overweight/underweight would have returned an excess of 0.26%/yr less trading costs.

[scb175]

I thought this link would fit in with the topic. A must see for those interested in this conversation.

http://www.trade-ideas.com/StockInfo/_T ... onfig.html

[retired at 48]

Hi Folks

I'm en route transitioning to south Florida, and see my name is featured in JW's intro and discussion, so it is appropriate I respond with a few points.

First, I use the 200 day moving average with a buy in technique I called Pyramiding Up, best discussed in this thread:

http://www.bogleheads.org/forum/viewtop ... vii+photos

This was my trial and error adaptation to investing with 100% mutual funds in the 1970's, and on.

Secondly, there is a worthwhile short video on looking at moving averages historically, here:

http://www.youtube.com/watch?v=bN9WUIXaRr4

Some other pointers.

I am not a short term trader/timer. I have cautioned the 200 dayma works better with buys, than sells. Tendencies exist to get whipsawesd on sells.

It is encouraging to see that many of you have backtested this and got positive results versus buy and hold returns.

TheEternalVortex asked "why 200 days?...why not other time periods. During the 1970's I only saw the data from a mutual fund service that had 200 day moving averages. So one worked with what one had. Remember, no computers in those investing times. Could there be a better number of days? Certainly. But here's the key:

JW Nearly Retired referred too this as a technical indicator. Here, I partly disagree. I have studied loads of tech indicators in my time...none work well. I personally view the 200 dayma as not a cause/effect indicator, but rather a mathematical smoothing of the data. The reason to do this smoothing is to identify when an uptrend or downtrend is underway. The 200 day ma exists for all funds, indices and markets.

One thing can be said with 100% certainty...No bull market will ensue without the NAV crossing the 200 dayma on the upside.

Those of you who backtested and noted there would be no buying in 2008, any month, are indeed correct. My pyramid up technique yielded the same...NO BUYING.

Some will call this lucky. I don't claim predictive powers. But it is demonstrated these are the facts to date, and have been posted along the way, as such.

I've invested for forty years now. My learning is that I should have payed more attention to 200 day moving averages, not less. Especially regarding international funds, this year, as I took too little off the table.

A note of thanks to JW Nearly Retired who, skeptical at first, backtested this to various funds, and saw the positive results. Most Bogleheads have learned that "market timing" is impossible...so any such claims face a headwind on this forum. Fair enough. But used in long term investing strategies to take some off the table at times of greed, adding it back in in times of fear (like now) pay off in the long run. Use of 200 day moving averages helped me greatly determine trend changes.

retired at 48

[JW-Retired]

Try back-testing it against the S&P 500 and make sure to include the full 40 years starting from, say, 1968. Timing looks good when there are relatively long multi-year trends to follow - it looks less good when there's a lot of choppy up-and-down actin.

OK, went back to 1952,
Start date= 1952.8350, price= 24.60, Portfolio Value= 10,000.00
Last exit date= 2007.9535, price= 1465.05, system Portfolio Value= 466,337.42, total trades since 1952= 136
Recent 11/7/08 Values: price= 930.99, system Portfolio value= 466,337.42, buy-hold value= 378,451.22

This seems less than stunning to me. The system got you out in Dec 2007 and has you ahead of buy-hold right now. Most of the time you are not as good as buy-hold, but it is crash insurance. Maybe fiddling with the averaging period and/or the tolerance bands might make it better. I will try that.

Thanks for all the good comments by all.
JW

[img][URL=http://imageshack.us][img]http://img50.imageshack.us/img50/2627/s ... em2hx0.gif[/img]

[JW-Retired]

There is no fool proof system for market timing. If there was someone would have figured it out by now and everyone would use it.

I always thought that too. However, Retired at 48 pointed out that only small investors can make any useful use of timing. Large funds who try to sell out the next day after a timing signal would simply be unable to. The huge orders would crash the price. We small fry can.
JW

[JW-Retired]

done some Vanguard stuff and get this:
http://www.gummy-stuff.org/200-day-MA.htm#DP
For example:

Thanks, excellent stuff for me to check if my own simulations are working right.
JW

[JW-Retired]

The point of the newsletter was using a 39 week moving average, almost the same as the 200 day moving average. It looked good on paper and would "save" the investor from terrible bear markets.

Unfortunately in the real world there were terrible whipsaws. This was before ETFs so you got end of day mutual fund pricing which made matters worse. So instead of death by the guillotine of a bear market, it was death by a thousand cuts of whipsaws and constantly being on the wrong side of the market (out when markets rose and in when markets started going down).

In the little I have done with this it is clear that the in/out churning in a noisy market that has no real direction is the big problem. The timing system keeps selling a little low and buying a little high over and over. But if you stop until there is direction, when do you start up again? Just maybe we can think up something that works. Anyway, it's entertaining to work on it.
JW

[DP]

Hi,

In the little I have done with this it is clear that the in/out churning in a noisy market that has no real direction is the big problem. The timing system keeps selling a little low and buying a little high over and over. But if you stop until there is direction, when do you start up again? Just maybe we can think up something that works

FYI, in my tests on a diversified set of funds I get:
ST LT AR HoldLen DrawDown
Buy&Hold 4.00% 2652 -53.55%
1 200 6.95% 32 -21.48%
50 200 8.22% 236 -19.18%
10 month 7.11% 188 -23.91%

Hold length is the number of days in the test divided by the number of trades. Whipsaws cannot be eliminated, but they can be reduced either with the use of moving average cross overs or only trading against the moving average monthly .. and both actually increased returns.

Don

[inve$t0r]

Beware of geeks bearing formulas. You have a huge recency bias if you're using the last 200 days, for obvious reasons.

[DP]

Hi,

Beware of geeks bearing formulas. You have a huge recency bias if you're using the last 200 days, for obvious reasons.

I'm not clear on your point. If you'd read the paper referenced by Mebane Faber you would know that the 10 month sma was tested over an extensive history and the paper was written in 2006. So the current results of his method, which completely avoided the bear market, is completely out of sample.

Don

[rwwoods]

You can download the Faber paper, A Quantitative Approach to Tactical Asset Allocation, at this site:
http://papers.ssrn.com/sol3/papers.cfm? ... _id=962461

[psteinx]

I'd be interested in seeing a huge grid of analysis - overlaying many lag periods and different buffer sizes. Then I'd like to see the results for different markets, time frames, and perhaps different assumptions about transaction costs and taxes. Results would include simple returns, risk-adjusted returns, number of trades, and perhaps a few other things.

I'm confident if you backtest enough strategies and tinker with the variables enough, some would look good.

And in fact, it's possible that there's a bit of merit to these kinds of strategies, provided your costs are low enough, because from various things I've read, there is a bit of momentum to the stock market, short term.

But I wonder how robust these strategies are, on the whole. If we backtest and find that a 200 DMA with a 1% boundary looks good, but that the same thing with a 2% boundary does not, or that using a 180 DMA yields substantially different results, then the whole exercise becomes a bit suspicious. In theory, by varying a number of the key parameters, you could come up with thousands or perhaps even millions of potential strategies. If 70% of strategies with reasonable parameters work, then I'd be somewhat interested. If only 15% of them work, and somebody is cherry picking the most successful from within that 15% to advocate the strategy, then I'd be very skeptical. Of course, many folks might have different thresholds/standards for saying something "works", as well.

[cflannagan]

I'd be interested in seeing a huge grid of analysis - overlaying many lag periods and different buffer sizes. Then I'd like to see the results for different markets, time frames, and perhaps different assumptions about transaction costs and taxes. Results would include simple returns, risk-adjusted returns, number of trades, and perhaps a few other things.

I'm sure if you backtest enough strategies and tinker with the variables enough, some would look good.

And in fact, it's possible that there's a bit of merit to these kinds of strategies, provided your costs are low enough, because from various things I've read, there is a bit of momentum to the stock market, short term.

But I wonder how robust these strategies are, on the whole. If we backtest and find that a 200 DMA with a 1% boundary looks good, but that the same thing with a 2% boundary does not, or that using a 180 DMA yields substantially different results, then the whole exercise becomes a bit suspicious. In theory, by varying a number of the key parameters, you could come up with thousands or perhaps even millions of potential strategies. If 70% of strategies with reasonable parameters work, then I'd be somewhat interested. If only 15% of them work, and somebody is cherry picking the most successful from within that 15% to advocate the strategy, then I'd be very skeptical. Of course, many folks might have different thresholds/standards for saying something "works", as well.

Yep. Need to eliminate the risk that the discovery is actually "datamining" which is finding a solution that fits only for the given sample, even if the sample is our entire history, because it just happens to fit that, but does not necessarily mean it'll work in the future.

And also eliminate the risk of "crystal ball effect" - we have to remember we make trades after the fact, not before. Ie: when some 200MA line is crossed.. the trade normally should happen on the next day, not at the exact moment..

[retired at 48]

Need to eliminate the risk that the discovery is actually "datamining" which is finding a solution that fits only for the given sample, even if the sample is our entire history, because it just happens to fit that, but does not necessarily mean it'll work in the future.

Bogleheads, you see it is not likely one can ever meet the standards of backtesting and data verification some will require.

It is especially difficult when the standard could encompass all of history, but will still not suffice! Whew.

What is new in the last decade which makes both enhanced market timing and momentum investing possible is: Computers, which cascade tons of data into ready use, almost instantly; tax advantaged spaces, like IRA's, that permit the avoidance of capital gain taxes; very low cost drags, ranging from commission free trades to perhaps $8 per trade; and exchange traded funds which permit real time buying/selling, rather than waiting to market closing prices.

More on technical aspects to follow.

retired at 48 (R48)

[JW-Retired]

And also eliminate the risk of "crystal ball effect" - we have to remember we make trades after the fact, not before. Ie: when some 200MA line is crossed.. the trade normally should happen on the next day, not at the exact moment..

Yep, I have no crystal ball built into what I'm doing. If the close price signals something the computer trades at the next day's open price. So this simulation is something I could actually do, if I ever start to believe in it.
JW

[dmcmahon]

This seems less than stunning to me. The system got you out in Dec 2007 and has you ahead of buy-hold right now. Most of the time you are not as good as buy-hold, but it is crash insurance. Maybe fiddling with the averaging period and/or the tolerance bands might make it better. I will try that.

Actually, your results are about what I'd have hoped for and they are encouraging.

Here's why: people feel the pain of downdrafts more acutely than they feel the pleasure of upside. I've often wondered if it was possible to devise a timing system that lagged the market by a small amount but that eliminated the huge potholes. Such a strategy might, psychologically, be helpful to many.

One concern is of course tax side-effects - it would be useful to plot a line showing cumulative realized capital gains, if that's possible, to enable people to get a sense of what they might have paid out in taxes with the timing system.

It seems inevitable that anything short of perfection in timing is going to lag the market, especially after taxes are considered. Still, the question is interesting: would you like an investment that gave you the market return minus 1%, but that ensured that the worst-case draw-down was, say, 10%?

Might I suggest a tweak to the system? Try adding the 50-day moving average. Make the new sell signal a break below the 50-day when the 50-day is below the 200-day. Make the new buy signal a move above the 50-day when the 50-day is above the 200-day.

[psteinx]

retired at 48 -

It seems you're mocking the concerns about data mining, but they are, IMO, very legitimate.

It is relatively trivial to locate successful systems when backtesting enough candidates. But if a system truly has a logical underpinning and is sufficiently effective, it should be robust to out of sample testing and to tweaking of its core parameters.

If a given system works very well with a 200 DMA, but does not work well with a 197 DMA, then I would be highly suspicious that the results for the 200 DMA were a random artifact. The underlying logic, I think, is that momentum strategies work because of positive short term serial correlations, which in turn are caused by short term market under-reactions.

===

I agree that developments of the last decade or so make such strategies potentially more viable. OTOH, I wonder if these same developments may have substantially reduced or eliminated the potential market underreactions that theoretically supported the concept in the first place. Markets that are not very liquid, with high transaction costs, and where information is relatively slowly distributed seem like candidates for short term underreaction. But pretty much all markets that we would likely invest in have seen these issues diminish rapidly of late. Has the underreaction diminished too?

[Speedie]

OK, went back to 1952,
Start date= 1952.8350, price= 24.60, Portfolio Value= 10,000.00
Last exit date= 2007.9535, price= 1465.05, system Portfolio Value= 466,337.42, total trades since 1952= 136
Recent 11/7/08 Values: price= 930.99, system Portfolio value= 466,337.42, buy-hold value= 378,451.22

This seems less than stunning to me. The system got you out in Dec 2007 and has you ahead of buy-hold right now. Most of the time you are not as good as buy-hold, but it is crash insurance. Maybe fiddling with the averaging period and/or the tolerance bands might make it better. I will try that.

Thanks for all the good comments by all.
JW

Do your figures include any return on the cash that you would have been holding when the system had you out of the market?

[DP]

Hi,

It seems you're mocking the concerns about data mining, but they are, IMO, very legitimate.

They are very legitimate concerns, but I think there is already strong supporting evidence referenced in this thread. If you have read the referenced papers and that's not enough, and you are actually interested, you should test it yourself. That's what I did and the results are compelling and robust across all parameters I tested. And using either crossovers or monthly averages, average holding times approach 1 year, longer for winning trades.

It's definitely not all data mining, I saved a lot of money based on the 10month sma this fall ... and to answer a question so often raised on this forum in reference to market timing: I'll know exactly when to get back in. Wish I had used this method for more of my equity holdings.

Don

[rwwoods]

It has often been said that "Past performance does not predict future performance," but consider the following.

Why do you believe in using index funds vs managed funds? Probably because PAST PERFORMANCE has shown that it is effective.

Why do you believe in having a diversified portfolio? Probably because PAST PERFORMANCE has shown that it is effective.

Why do you believe in using the various asset classes you hold? Probably because PAST PERFORMANCE has shown that it is effective.

Why do you overweight certain areas such as small cap value? Probably because PAST PERFORMANCE has shown that it is effective.

Why do you believe that your passive, diversified portfolio will continue to provide effective long term performance in the future? Probably because PAST PERFORMANCE has been effective.

Then why do you NOT accept that PAST PERFORMANCE can be used to determine the effectiveness of a 200 day moving average? It has been demonstrated that it has reduced portfolio risk by about half as measured by standard deviation and maximum draw down while maintainting similar returns as a buy-and-hold strategy.

One has to admit that all investment strategies are based upon our limited knowledge of past performance. In creating any strategic plan, one must be careful on how to intrepret this knowledge. As has been noted, data mining can be harmful. Thus one must be careful in the use of techincal analysis. However, the 200 day moving average has been shown to be effective when properly applied.

Will any stratigic investment plan or tactic continue to be effective in the future? Obviously, no one knows. We can only draw upon our limited knowlege of past performance.

[JW-Retired]

Do your figures include any return on the cash that you would have been holding when the system had you out of the market?

Nope, no return on the cash and no account for cap gains/dividends on the fund when you are invested. Just assuming this is more or less a push because it's hard to do anything else.

I'd be interested in seeing a huge grid of analysis - overlaying many lag periods and different buffer sizes.

Huge it's not, but I did a little more with the SP500 data and it says there is a fairly broad moving avg period optimum that a period like 200 days is best. I actually get 207 days is optimal for this data, but I would not claim there is any generality to 207 days. In this case (GSPC) the 200 is at a little local minimum. I also adjusted the i/o tolerance band and find I get best results with 0.5%. See figure.

Using these cherry pickedish parameters (0.5% i/o tolerance & 207 days moving average) the system gives:
Start date= 1952.8350, price= 24.60, Portfolio Value= 10,000.00
Last exit date= 2007.9890, price= 1479.83, system Portfolio Value= 567036.48, total trades since 1952= 170
Recent 11/7/08 Values: price= 930.99, system Portfolio value= 567036.48, buy-hold value= 378,451.22

IMO, this makes the "system" worth looking at some more.

I will try to apply it to other funds & indexes. Hopefully more later this weekend.
thanks for all the great comments.
JW

[JW-Retired]

RE: I will try to apply it to other funds & indexes.

Used the mov_avg=207 that the SP500 likes and applied this to some more funds. Tried both 0.5% and 1% bands. Results follow: Hope the table is legible.
-------------207/0.5%--------------
Fund.....date-in...price-in..shares-in..invested..out-worth...buyhold...#trades
VEIEX.. 1996.5859 . 11.84 . 844.59 . 10000 .. 35673.37 .. 13116.55 ... 32
SPY...... 1994.4531 . 46.55 . 214.82 . 10000 .. 23930.00 .. 20163.27 ... 70
VEXMX.. 1992.5339 . 15.80 . 632.91 . 10000 .. 16931.98 .. 15632.91 ... 90
QQQQ... 2003.1978 . 25.73 . 388.65 . 10000 .. 13970.78 .. 12122.04 ... 34
DIA ..... 1999.8220. 106.06 . 94.29 . 10000 ... 7355.09 ... 8498.02 ... 72
XLE ..... 1999.9699 . 26.56 . 376.51 . 10000 .. 17358.16 .. 18768.83 ... 52
XLF .... 2000.2245 . 23.72 . 421.59 . 10000 .. 10295.14 ... 5994.94 ... 52
IYH .... 2001.8015 . 62.00 . 161.29 . 10000 .. 10035.86 ... 8320.97 ... 32
IYK ..... 2001.6208 . 42.00 . 238.10 . 10000 .. 10412.06 .. 10892.86 ... 44
TMW ... 2002.1739 . 82.45 . 121.29 . 10000 .. 11892.41 ... 7781.69 ... 34
................................................................_______________________
.......................................................totals.... $157855 .... $121,292 .....
GSPC....1952.8350..24.60....406.50....10000...567036.48...378451.22...170

------------207/1%------------------
Fund.....date-in...price-in..shares-in..invested..out-worth...buyhold...#trades
VEIEX.. 1996.5859 . 11.84 . 844.59 . 10000 .. 35876.01 .. 13116.55 ... 28
SPY...... 1994.4531 . 46.55 . 214.82 . 10000 .. 27258.45 .. 20163.27 ... 42
VEXMX.. 1992.5339 . 15.80 . 632.91 . 10000 .. 15379.57 .. 15632.91 ... 66
QQQQ... 2003.1978 . 25.73 . 388.65 . 10000 .. 14575.60 .. 12122.04 ... 20
DIA ..... 1999.8220. 106.06 . 94.29 . 10000 ... 6894.72 ... 8498.02 ... 54
XLE ..... 1999.9699 . 26.56 . 376.51 . 10000 .. 18667.39 .. 18768.83 ... 38
XLF .... 2000.2245 . 23.72 . 421.59 . 10000 .. 10611.86 ... 5994.94 ... 32
IYH .... 2001.8015 . 62.00 . 161.29 . 10000 .. 9876.37 ... 8320.97 ... 22
IYK ..... 2001.6208 . 42.00 . 238.10 . 10000 .. 9624.92 .. 10892.86 ... 36
TMW ... 2002.1739 . 82.45 . 121.29 . 10000 .. 11834.75 ... 7781.69 ... 22
..............................................................__________________________
......................................................totals.... $160,600 .....$121,292
GSPC ...1952.8350...24.60...406.50...10000...528074.98...378451.22...122

This system worked pretty well on TMW & SPY and great for VEIEX, poor for DIA. Did better than buy-and-hold on 7 of 10 funds. On avg over the 10 funds it's up about 30% compared to buy and hold.

Might I suggest a tweak to the system? Try adding the 50-day moving average. Make the new sell signal a break below the 50-day when the 50-day is below the 200-day. Make the new buy signal a move above the 50-day when the 50-day is above the 200-day.

OK, tried it but it didn't work well. Waiting until the 50 day avg moves gets you in and out too slow compared to NAV cross. It does reduce the trades by a lot. Results on SP500 below:
---dmcmahon idea----207/0.5%/50-200day cross-------------
GSPC 0.5%/50 1952.8350 24.60 406.50 10000 393805.65 378451.22 54
---dmcmahon idea----207/0.0%/50-200day cross-------------
GSPC 0.0%/50 1952.8350 24.60 406.50 10000 418140.12 378451.22 54

I'm done with this for a while. But I think I will try it in a small way when the market allows.
JW

[retired at 48]

dmcmahon wrote, partly

Actually, your results are about what I'd have hoped for and they are encouraging.

Here's why: people feel the pain of downdrafts more acutely than they feel the pleasure of upside. I've often wondered if it was possible to devise a timing system that lagged the market by a small amount but that eliminated the huge potholes. Such a strategy might, psychologically, be helpful to many.

dmcmahon, to me this is the most significant observation. My pyramid Up technique partly evolved in trying to put to action Warren Buffett's words like "Rule #1: Don't lose money; Rule#2: Read rule 1." But what was one to do?

You are right in that the 200 day ma optomizes on avoiding large losses, sacraficing a little on the upside to achieve this. With Pyr Up buying in increments, averaging up instead of down, it further adds this insurance policy. If one can avoid market losses, then one will gain.

Imagine that during the 1970's one had about 25 no load funds, all actively managed. One treaded these waters carefully. The Pyr Up technique/200 day ma was my method.

R48

[retired at 48]

retired at 48 -

It seems you're mocking the concerns about data mining, but they are, IMO, very legitimate.

It is relatively trivial to locate successful systems when backtesting enough candidates. But if a system truly has a logical underpinning and is sufficiently effective, it should be robust to out of sample testing and to tweaking of its core parameters.

If a given system works very well with a 200 DMA, but does not work well with a 197 DMA, then I would be highly suspicious that the results for the 200 DMA were a random artifact. The underlying logic, I think, is that momentum strategies work because of positive short term serial correlations, which in turn are caused by short term market under-reactions.

No...I don't want to mock concerns on data mining, but when one says the entire history of data is not enough, then what is one to do. rwwoods's response above porbably sums it up. If one is looking for 100% guarantees, then it is unlikely.

It is encouraging that JW Nearly Retired ran data showing 207 days to be optimal...close enough to 200. And that results were robust about these total days selected.

The goal is to be investing in upward moving, or bull markets. Then, it is the most unfortunate individual who invests and on that day the market turns. But it can happen. That's why the Pyramid Up technique of investing over an average of five, all upwardly moving price buys, further mitigates against loss. One is in a position to act with "losing gains" versus "losing money" in a subsequent downtrend move.

R48

[cflannagan]

Jumping back in kinda late without reading all the comments, so if this was already mentinoed.. sorry 'bout that.

Looks like 200MA works quite well for preventing large losses from last peak. I also observed this independently and before this thread began.

But unfortunately, it also often misses large upward movements while the 200MA has not yet crossed up past the NAV.

So maybe we can tweak the system for jumping back in, while keeping the 200MA rule for getting out.

[retired at 48]

retired at 48 -


I agree that developments of the last decade or so make such strategies potentially more viable. OTOH, I wonder if these same developments may have substantially reduced or eliminated the potential market underreactions that theoretically supported the concept in the first place. Markets that are not very liquid, with high transaction costs, and where information is relatively slowly distributed seem like candidates for short term underreaction. But pretty much all markets that we would likely invest in have seen these issues diminish rapidly of late. Has the underreaction diminished too?

Hi...can you further define "underreactions." I want to comment properly, but want to be on same page. Are you referring to "has this been incorporated into the market?" OR "what do you mean by theoretically supported the concept in the first place?" Thanks.

R48

[retired at 48]

JW Nearly Retired...again, thanks for running the data. I am an old timer and have some difficulty doing that.

Also, the proof is in my head

Two other points.

First, I never planned that one would simply take an entire portfolio to cash, timing back and forth...but it is the way to backtest the data. The results are very impressive. To me, combining the 200 day ma with "valuations matter" themes, and peoples psychological needs when investing, results in the Pyr Up method.

Second, D. Swensen of Yale says investing to him is mostly science but also art. The art part is "when to invest when the NAV crosses the 200 day moving average trend line." Your use of 1% and 1/2% is useful data, but note one cannot invest at those precise points.

I tried to have a robust day in crossing the 200 day ma before taking action. And therein lies the "art" that some can question. There is no easy substitute for experience here...feel if you will. But I definitely sense changes in direction and market robustness, up or down.

retired at 48 (R48)

[Adrian Nenu]

We need to simply remember that downside violations of 200 month moving averages are quite the rare beasts. As we suggested when we last covered this topic, the only times we’ve seen occurrences where this has happened have been accompanied by periods of very meaningful fundamental economic challenge. As you can see below, we’re there again.

So, either the forced liquidation and deleveraging of the moment is anomalistic enough that it is not allowing the financial markets to more properly discount forward economic reality to come in current prices, or this very deleveraging process itself will further contribute to negative real world economic outcomes ahead and the markets are suggesting as much.

Sounds like a version of DCAD (dollar cost averaging down), yours based on 200 day moving avg, the other more familiar to me based on the inverted yield curve.

Adrian
anenu@tampabay.rr.com

[retired at 48]

We need to simply remember that downside violations of 200 month moving averages are quite the rare beasts. As we suggested when we last covered this topic, the only times we’ve seen occurrences where this has happened have been accompanied by periods of very meaningful fundamental economic challenge. As you can see below, we’re there again.

So, either the forced liquidation and deleveraging of the moment is anomalistic enough that it is not allowing the financial markets to more properly discount forward economic reality to come in current prices, or this very deleveraging process itself will further contribute to negative real world economic outcomes ahead and the markets are suggesting as much.
[/quote](james 22)

Sounds like a version of DCAD (dollar cost averaging down), yours based on 200 day moving avg, the other more familiar to me based on the inverted yield curve.

Adrian
anenu@tampabay.rr.com

james22, I'm not so sure what you refer to by stating downside violations are "rare beasts"? My experience, confirmed by JW's analysis, is they are far more frequent. In fact, whipsaws are somewhat of a concern.

But the beauty of the market is that probably 85% of the time the NAV is trending above the 200 day ma, thus one should be comfortable being invested. One wants to be in during bull phases. That sell signals, which can be used as "taking some off the table" are fewer is good.

Here's an example. If one is about to rebalance a portfolio, and the fund one is contemplating buying has the NAV trending below the 200 day ma (like this year), then wait...patience...rebalance the day it crosses, on the upside.

Recent Bogle detailed studies show no gain/loss due to rebalancing. I suggest rebalancing using 200 day moving average guidance, including times the NAV greatly varies from the 200 day ma (return to the mean themes), will add alpha.

Those who are waiting this year will certainly add alpha.

R48

[DP]

Hi,
Interesting that JW's results show better results with the simple 200 ma than a 50 x 200 ma crossover. My tests on 6 funds over the past 10.5 years (results posted previously) do not confirm this. Perhaps simply a matter of the timeframe and variance of results. In any case I find the number of trades in the simple moving average too high for my liking. A moving average crossover, even if it's something like the 10 or 20 vs. 200 significantly reduces trades and still performs very well. And the monthly moving averages are suprisingly good in the long run.

So you held until now and have absorbed 30-40% loss on your ETF portfolio. Now you "see the light" and decide to switch your strategy to using the 200 d MA to determine exit and entry points. But all your ETFs are way below the 200 d MA, so what do you do? You sell everything and wait until the 200 d MA is above the current price to buy back or you decide that it is too late to switch and ride the market? Erwin

So maybe we can tweak the system for jumping back in, while keeping the 200MA rule for getting out.

This raises an interesting point. If one was following a buy and hold strategy, and IF we are near the bottom, now would not be a good time to sell and switch to a long term moving average. I've seen a slightly more aggressive strategy which buys whenever the price is above the 50 OR 200 moving average. This strategy assumes the market is bullish more than bearish, which is true in the long run.

I tested using simple price vs. the average and when the 10 ema was above the 50 or 200 ema. The results for the latter were about as good as any combination I tested, but still I concluded that it added little value over a simple 50 x 200 ema. In thinking about the question, though, if one were to start a timing system now, this system might have an edge because it would get you back in quicker, as would similar variations.

Don

[james22]

james22, I'm not so sure what you refer to by stating downside violations are "rare beasts"? My experience, confirmed by JW's analysis, is they are far more frequent. In fact, whipsaws are somewhat of a concern.

200 month ma, R48.

"...there is an old truism in the markets that in very severe equity bear episodes, the 200 month moving average of the equity indices is a potential downside target."

http://contraryinvestor.com/2008archives/moaug08.htm

[JW-Retired]

Interesting that JW's results show better results with the simple 200 ma than a 50 x 200 ma crossover. My tests on 6 funds over the past 10.5 years (results posted previously) do not confirm this.

DP,
I have to admit I gave up too easy on the 50x200 after just testing it on the 50 years of SP500. After your comments above I thought I'd better check it on all the 10 funds done earlier with the NAV crossing system. It works much better on this fund group and gives better results than the NAV. Far fewer trades. I have no idea why it is not so hot with GSPC.
50/207ma system
Fund..date-in...price-in..shares-in..invested..out-worth...buyhold...#trades
VEIEX...1996.5859....11.84....844.59...10000....37894.21....13116.55....12
SPY.......1994.4531....46.55....214.82...10000....34717.84....20163.27....12
VEXMX...1992.5339....15.80....632.91...10000....20071.43....15632.91....24
QQQQ....2003.1978....25.73....388.65...10000....12196.57....12122.04.....8
DIA.....1999.8220...106.06.... 94.29...10000.....8297.71.....8498.02....20
XLE.....1999.8576....27.34....365.76...10000....22062.31....18233.36....10
XLF.... 2000.2245 .. 23.72 .. 421.59.. 10000 .. .9039.49 ....5994.94....16
IYH.... 2001.8015 .. 62.00 .. 161.29.. 10000 .. 10281.85 ....8320.97....10
IYK.....2001.6208....42.00....238.10...10000....11940.27....10892.86....10
TMW.....2002.1739....82.45....121.29...10000....13820.64.....7781.69.....8
..................................................................._______________________
...........................................................totals.... $180322 .... $121,292
GSPC...1952.8350...24.60...406.50...10000......418140.12...378451.22.....54
thanks for the experienced comment DP.
JW