Smt 2/222

[34rlsa]

Recently in a few different threads the idea of using the 200 day moving average to time the market have been discussed. From a long term perceptive it does not appear to be an effective way to invest. However for those in retirement is there any merit to utilizing a similar strategy?

I came across this arcticle at I-orp.com and found it interesting. I did a quick search and did not find it mentioned here on the forum.

Simple Market Timing

Using a 1 month chart you plot the 2 day moving average and the 222 day moving average. When the lines cross you buy or sell accordingly.

http://i-orp.com/modeldescription/smt%5B1%5D.pdf

My understanding from the arcticle... its not so much about market timing to maximize returns but more about handling volatility during retirement to better control withdrawal rates.

For the risk averse who may not have the fortitude to adhere to a buy and hold strategy... is this a practical alternative you have used or would consider using?

 

[34rlsa]

No replies?

Is Welch's method completely worthless? Even in a non taxable account?

 

[Lakewood90712]

No replies?

Is Welch's method completely worthless? Even in a non taxable account?

I think you can guess the answer.

I am a market timer. Don't use charts , just use the Warren Buffet adage :
" Be fearful when others are greedy , and be greedy when others are fearful ". I am "out of the market" equity wise for long periods. I buy quality , large cap individual companies when panic has turned to capitulation (2009-2010), and sell when the talking heads at CNBC , and elsewhere are bullish. Far from Ideal , but I am a worrier by nature, and still sleep soundly with this crude method. Charts and "analysis of the "Market" is useless IMO , Mega-trends, and Mega-events move the markets.

I don't recommend it, but it works for me.

 

[pb4uski]

I seem to recall back-testing something like this at one time (I think I used VTSAX rather than the S&P 500) and it was marginally better as I recall, but in retirement I have better things to do and like the set it and forget it of buy, hold and rebalance.

 

[target2019]

No replies?

Is Welch's method completely worthless? Even in a non taxable account?

The use of long and short signals is not new, as the author states in the article. His 2/222 method tests well according to historical data. Now, place real money into an experiment and track results. Set up a method to track new data and compute the stats. Also, establish a barometer to compare your results to buy and hold and other methods.

 

[Lsbcal]

If I were to take up one simple timing approach I'd consider a monthly moving average type like discussed here: Timing Model – Meb Faber Research – Stock Market and Investing Blog

One could use this to move a portion of one's portfolio out of the market so as to avoid periods like 2002 and 2008 that might have grown into a period like the 1930's. So if one was 60/40 maybe they would go to 30/70 on such a signal. There would be whipsaws and one should know the history of such approaches. No sense in adapting this if you think it is without some pain.

I do have a personal methodology but it's not based on simple moving averages.

 

[photoguy]

Keep in mind the risk of data mining or multiple hypothesis testing. The author is probably testing out ~1000 different models and selecting the best one.

The 1000 come from at least 50 different days for the long term moving average (190 to 240) and probably maybe 20 or so options for the short term to decide on 2 day moving average (he doesn't actually say how many variations he considered for the short term MA but let's say he considered 1-20).

I wonder why all of these methods are reported as backtesting instead of I followed this method for 30 years and actually obtained the results in this paper?

 

[Lsbcal]

It is quite easy to do a sensitivity analysis. If a model is highly sensitive to parameter variations, it should be rejected.

 

[photoguy]

Totally agree that the author should have done a sensitivity analysis. He also should have done an out-of-sample test but he didn't do either.

 

[CaliforniaMan]

Keep in mind the risk of data mining or multiple hypothesis testing. The author is probably testing out ~1000 different models and selecting the best one.

The 1000 come from at least 50 different days for the long term moving average (190 to 240) and probably maybe 20 or so options for the short term to decide on 2 day moving average (he doesn't actually say how many variations he considered for the short term MA but let's say he considered 1-20).

I wonder why all of these methods are reported as backtesting instead of I followed this method for 30 years and actually obtained the results in this paper?

+1000 Yes this!

 

[samclem]

Somewhat OT: We spent a lot of time in school learning calculus. But in my real life, the 35+ years after HS, statistics (and probability) have been much more useful. I've seldom actually needed to work a statistics calculation, but just having the concepts in mind are enough. It's a pity that "mathematical reasoning for real life" isn't covered more thoroughly in high school. It could do a lot of people a lot of good, the actual calculations and steps aren't very hard, and to an impatient HS student, the tangible utility (you don't have to stretch to think of a realistic word problem in these areas) is a plus.

 

[CaliforniaMan]

What I think is that most people don't realize how very easy its is to put together, using various regression techniques, a stock market prediction that will beat buy and hold. You can use various moving averages or other smoothing filters, sine waves, economic indicators, even random series. The important thing is that the model does not even have to make sense (using the price of figs in Afghanistan, or biscuits sold in a Londonderry cafe, skirt heights or Superbowl winners, or your wife's coin flips). Of course since your model "works" you can always make an explanation for why the fig price is important. Many decades ago when I realized I could do this, I thought hey, I could make money doing this, fooling people I could predict the market. After all people really want to believe anyway, so it is not all that difficult. No, I didn't do it, lots of things you know how to do you shouldn't do, this was one of them for me. But it is really easy. Photoguy and samclem make very good points. And when you see a number like 222, be suspicious, very suspicious. Of course common sense tells you this cannot work anyway. If it did, everyone would use it, then it wouldn't work. Therefore it cannot work.

Now it is not that people are always trying to fool you, very often, maybe even most of the time they believe it themselves. I have one example in biology. We had population estimates of maybe a hundred or more species of plankton (microscopic ocean critters) through time from core samples. We wanted to find their relationships if any. We found a lot of significant correlations, some made good sense, others were strange, but all were interesting and was the making of a good paper. I was young at the time and only starting playing with this kind of thing. I had the idea to correct the correlations for the number of comparisons. Oops. All of the significance disappeared. We really had just about what you would expect by chance. Remember you cannot correlate time series (the chance correlations tend toward +/- 1.0.) Did anyone want to hear about this finding? Nope.

At any rate it is fun to try and I am sure many of us here over the years have put together various regression or Fourier projections of the market. Unfortunately if you are honest with yourself (as you always are after enough losses) you realize you are just fitting to the data. It tells you nothing about the future. Your results are meaningless. Better to spend your time doing something else more productive or more fun. Hey, unless you want to trick people and sell newsletters.

 

[34rlsa]

I don't disagree with any of the replies... but my question wasn't about BEATING buy and hold with a particular timing strategy. Rather... Is there any merit to using a similar strategy in retirement to avoid severe declines to the value of your portfolio to the point where the value may no longer support your withdrawal rate.

Say you retired in 2000...

What would be interesting to see is the authors same 2/222 method (realizing there is nothing magical about using these particular moving averages ) and the difference between "buy and hold" verses "timing" while taking a 4% withdrawal from a 1M dollar portfolio over the last 14 years.

 

[unclemick]

Speaking left handedly. If one steps back and looks at my Target Retirement Portfolio market timing has been handed over to the 'impersonal non-emotional Vanguard Computers' shifting automatically(aka rebalancing to a sliding mix of assets as I age).

Setting my male hormones aside, I am conceding their research ability is better than mine. After 40 years. (1966-2006).

heh heh heh - right now watching the Royals(Whaaa ) and channel surfing the Saint's. Mad money is for a few good stocks (under 5%) of total - if I hit the next Dell/Apple/unnamed super duper, I will become swell. But I am also not holding breath. Go Royals! Go Saints!

P.S. Staying the course during the last big dip and letting the computers rebalance was chewy but I did it. However - near the market lows I did throttle back my withdrawal rate for a year or so.

 

[springnr]

....... Now it is not that people are always trying to fool you, very often, maybe even most of the time they believe it themselves. I have one example in biology. We had population estimates of maybe a hundred or more species of plankton (microscopic ocean critters) through time from core samples. We wanted to find their relationships if any. We found a lot of significant correlations, some made good sense, others were strange, but all were interesting and was the making of a good paper. I was young at the time and only starting playing with this kind of thing. I had the idea to correct the correlations for the number of comparisons. Oops. All of the significance disappeared. We really had just about what you would expect by chance. Remember you cannot correlate time series (the chance correlations tend toward +/- 1.0.) Did anyone want to hear about this finding? Nope.

This years BAH Fest East was just held.
(This is the live stream, jump ahead 25 min to the real start)

 

[photoguy]

I don't disagree with any of the replies... but my question wasn't about BEATING buy and hold with a particular timing strategy.

The issues with the article's methodology affect all of the conclusions drawn from it which include the average return and maximum drawdowns.

Rather... Is there any merit to using a similar strategy in retirement to avoid severe declines to the value of your portfolio to the point where the value may no longer support your withdrawal rate.

A market timing strategy certainly could result in much better outcomes both in terms of max drawdown and average return. Personally I would be most interested in market timing strategies based on valuation. But I don't think that specific paper presented a particularly compelling strategy nor convincing evidence that it would work out of sample.

Say you retired in 2000...

What would be interesting to see is the authors same 2/222 method (realizing there is nothing magical about using these particular moving averages ) and the difference between "buy and hold" verses "timing" while taking a 4% withdrawal from a 1M dollar portfolio over the last 14 years.

Jim Otar's book (Unveiling the retirement myth) looks at a similar form of market timing in the context of retirement withdrawals. So you may want to look at that. However, as I recall the analysis in Otar's book was not particularly well done either (I wouldn't trust it but YMMV).

The author of the original paper appears to be the author of i-orp which I understand is a retirement planning software (is it like firecalc but also covers taxes?). Does i-orp not include this market timing method as an option?

 

[Lsbcal]

I don't disagree with any of the replies... but my question wasn't about BEATING buy and hold with a particular timing strategy. Rather... Is there any merit to using a similar strategy in retirement to avoid severe declines to the value of your portfolio to the point where the value may no longer support your withdrawal rate.

Say you retired in 2000...

What would be interesting to see is the authors same 2/222 method (realizing there is nothing magical about using these particular moving averages ) and the difference between "buy and hold" verses "timing" while taking a 4% withdrawal from a 1M dollar portfolio over the last 14 years.

If you can handle Excel or something similar, then you could do this analysis and let us know what you learn. I wouldn't want to invest my money on a study without all the details. Absent a willingness to do the hard work, I think it's best to concentrate on a simpler methodology.

You might just look at the current month to see what that would look like. I think you will find a sell and buy this October as the SP500 went above and below the 200 day SMA. Note if things go as they have a monthly plan will not give a sell at the end of October.

Also there are details that may not be evident on first glance. For instance, if a lot of trading is done on a daily basis then the entry points could come into play to raise or lower returns. If the idea is to sell when the price falls below the SMA, then are you dealing with fast moving intraday prices in the SP500? I still think a monthly technique is the way to go if one wants and SMA approach. See the paper I mentioned above as a starting point and then do your own analysis.

 

[34rlsa]

If you can handle Excel or something similar, then you could do this analysis and let us know what you learn. I wouldn't want to invest my money on a study without all the details. Absent a willingness to do the hard work, I think it's best to concentrate on a simpler methodology.

I have no idea how to run the analysis in Excel.

That's why I come here and ask questions

And thanks for the link to "A Quantitative Approach to Tactical Asset Allocation".

I agree a monthly approach would be much easier to implement and probably just as effective.

 

[Lsbcal]

In the past I investigated the 10 month SMA method. I ran an out of sample test using a simulated SP500 derived from the French Fama data. This simulated data matched very well with actual data for 1950 to 2008. So it gives me confidence in the entire time span from 1928 to 2008. Also using Dow Jones data results were pretty close.

Here is a table of the results. The first line reads from Jan 1928 to Jan 1938 the buy-hold CAGR (compound annual gain) was -5.2% and the 10mo SMA CAGR was 6.4%. The green indicates the winner for that period.

Sometimes buy-hold beats the 10 month SMA but over the entire time span the 10 month SMA was better. This moving average technique did avoid major declines like the Great Depression but in the WW2 era the buy-hold although scary would have provided much better returns.

Here is a table from actual SP500 data + dividends to get some more perspective:

So again, comparisons depend on the end point. If one using the 10mo SMA they have chosen to protect against major downturns and may give up some gains really good markets (the occasional whipsaw). Best in a retirement account.