World Beta Blog

I would like to see something presented which shows a chart of the rolling 12 month returns presented on a monthly basis. This is not a cumulative gain chart but a plot of the absolute return achieved over the previous 12 month period. I hope the information will show that the likelyhood of a positive return is very high.

I would love to see a long short version; a model that buys an inverse or leveraged inverse ETF instead of moving to cash. I've only glanced over the paper but I seem to recall that the time spent in cash is generally/relatively short-term.

I would also like to see the results of using ETF's with built in leverage (2x) v. margin, specifically to be able to use in retirement accounts.

What I would like to see is an backtest with a greater universe of ETFs to choose from (somewhere between 20 and 100 ETFs); e.g. the ETFdb 60 Index (http://etfdb.com/etfdb-60/) or a hand chosen list from a previous post here (can't find it, was a list with ~40 stock market sector/country/size ETFs, various commodity, bond ETFs, and alternatives likes VXX, DBV, etc...). The idea is to capture specific trends (currently e.g. Gold, Healthcare, China, ...) instead of broad markets.

Money is put into the top ~20% "momentum" ETFs (when above 200 MA).

"Momentum" means something like 3, 6, 12 month average performance divided by volatility.

One more try on the Links!!

ETFs with the 5,120,1 PPO

ETFs with the 5,220,1 PPO

IMHO, Choosing a large number of etfs will lead to a great deal of overlap and continual correlation. The real problem with that is that there is really not much price change difference between ILF and EWZ and so on with quite a few others. Dropping EWZ to move to ILF can be counter productive as the wiggle you are chasing will often reverse and have the other outperforming in the next period. Of course, we learned last year that when the stuff hits the fan, all assets were correlated. Even then the QTAA strategy did not result in a large and threatening decline.

I have looked at how the link for the stockcharts charts came out in the above post suggesting using the PPO of 5,120,1 and 5,220,1 and checking them weekly rather than monthly. The ETFs came through but the indicators did not. At the bottom of the charts is an option to choose an indicator. Use that so that the PPO is included.
Another attempt at the links:
(http://stockcharts.com/charts/candleglance.php?...)
(http://stockcharts.com/charts/candleglance.php?...)

I'm not too happy with any of the commodity trackers as they all have the problems with rolling the contracts. A recent introduction of an ETF that may avoid that is CRBQ even though it is made up of shares rather than pure commodities. Not enough time since introduction to know for sure how it will track the CRB.

I've only read the paper and not the book. I have ordered four copies to give to my children as mind stimulating gifts. I'll read it before they get it. I hope they pay attention.

I would like to note a couple of "problems" with the scheme. As described and presented with the results, the returns can lag the stock market returns for years at a time. The BIG players can overcome this with very cheap leverage which is not available to a normal small retail investor. After noting the problem. I will note that the likely returns for a year are positive and much better that what dcould have been achieved with any guaranteed type investment. that makes it very attractive for many investors and I can agree with the assessment.

The charts presented begin at a favorable time for the strategy as the markets tanked right at the beginning and the strategy avoided much of that. For the next 50 years or so the general stock market did better and finally caught up. That means that a human investor would have had to endure that underperformance and stick with the strategy until the 2000 - 2002 decline to again be in the lead and have a gain vs just owning the market equity index. Following that event the strategy underperformed again leading up to the 2008 crash where it again proved its value. I simply believe that it will be very hard for investors to hold to a strategy which continues to underperform for such long periods, waiting for the big crash to bring everyone else back to them in returns. People in their 30's or 40's investing today are likely to live to be 120 to 150 years old. It is pretty implausible to assume that there will not be future declines in the next 50 years or 75 years that will be as damaging as the ones of the past. However human nature will continue to draw us to whatever strategy is performing better in the short run, particularly after enduring several years of underperformance. It is a dangerous thing.

I have not backtested this but continue to BELIEVE that the strategy does not do very much to take advantage of the crashes in the various assets. It waits a bit long to get invested when the price moves up. Suppose is at 60 and 10 ma is 55. We all get out at 54 on the way down. Price eventually goes to a 25 low. If I buy in at 30 while someone else waits for 40 and then it goes back to 60 I have a 100% gain vs a 50% gain for those who delayed. We've both had positive returns but the difference in getting in maybe a month or two earlier is pretty dramatic.

My attempt at implementing the strategy uses a few more slices of the assets and follows:

Portfolio allocations with a modest anti dollar bias

US Stocks 20%
VTI 20%

Foreign Stocks 30%
VGK 10%
EPP 10%
VWO 10%

Bonds 20%
BND 10%
BWX 10%

Real Estate 15%
RWX 7.5%
VNQ 7.5%

Commodities 15%
DJP 15%

Then I've made additional changes to the monthly check up vs the 10 month average.
The price on one particualr day of the month is not necessarily a "good" price. I am using a 5 day ema. Further I will compare it to the 220 day ema (about 10 month) and the 120 day ema (about six months) and if it is greater than either of those I am IN. I am also using a weekly update / status check for trading.

It is pretty simple to implement today using the information on Stockcharts. Ignore the lines on the chart and look only at the PPO chart at the bottom:

120 day get in if PPO > 0
http://stockcharts.com/charts/candleglance.php?...

220 day
http://stockcharts.com/charts/candleglance.php?...

@aaCharley:

It would be interesting to see actual trading profits/losses generated by your system:
- with different starting dates (e.g. 10 years ago, 5 years ago, 3 years ago, 1 year ago),
- with different length of time staying invested (e.g. starting 10 years ago, show P/L after 1 year, 3 years, 5 years, 10 years).

Also, your system will need to make an assumption about the buying and selling price you will use, since we connot know at what intraday price the system would have bought/sold in the real world:
- if you assume a trigger signal can be observed at the end of the day, then you could assume buying at the opening price of the next day
- if you assume a trigger signal can be observed at the start of the day, the you could assume buying at the closing price of the current day

Furthermore, you will need to take into account trading costs (e.g. assume a fixed percentage of the order total value).

Also, you would need to make an assumption about the intrest you would receive on cash when your system is our of the market for one or more assets.

I would expect to only check the status on a weekly basis. And I would not assume any predetermined trigger signal that would execute during the week. One appeal of the QTAA approach is the simplicity and infrequency of trading. I would not expect that looking at two sets of charts over the weekend would be overly taxing.

Whatever the trading costs are, I would not expect them to be much increased from that of the original presentation. The whipsaws around the 120 day average would not be particularly different from the ones around the 10 month moving average. Of course, checking weekly rather than only once a month could increae them slightly but I would doubt that the costs would be too significant. It is possible that the smoothing from comparing the 5 day ema would in fact reduce those costs.

@aaCharley:

What would be truly interesting would be to see results of actual P/L calculations, in addition to the hypotheses being formulated here.

In particular, it would be interesting to see actual trading profits/losses generated by Meb's system, or your suggested enhancelent of that system:
- with different starting dates (e.g. 10 years ago, 5 years ago, 3 years ago, 1 year ago, et al),
- with different length of time staying invested (e.g. starting 10 years ago, show P/L after 1 year, 3 years, 5 years, 10 years; and the same for starting 1, 2, 3, 4, 5, 6, 7, 8, 9, 11, ... years ago).

If I am not mistaken, Meb's book (or white paper) does not show results for this kind of analysis of different starting dates, and lengths of investment. Although it is claimed that the method does work (= better returns, lower volatility) in other periods than the particular starting date (*), I have not yet been able to read evidence of this.

(*) As you also mentioned, Meb's analysis begins at a favorable time for his strategy as the markets dropped steeply shortly after the beginning of the period. It would therefore be interesting to see tables with the results of P/L calculations with other starting dates and different lengths of time invested.

General Comment:
Please don't add further filters, purchasing criteria, or DCA "steps". For me the appeal of this system is the simplicity. As you've stated before, this system isn't designed to beat the market but to protect capital. The TAA is a wonderfully simple means to do that.

Requests:
Discussion on alternatives for the less common investments in retirement plans. We don't have a commodity or currency fund in our 401k. Could an Energy Sector mutual fund act as a proxy for commodities?

Discussion on how to implement the broader portfolios described in your book using TAA. Obviously going with 3 assets when 10 or 20 are looked at can push you into a non-diversified position. Would you take 3/5ths of the 10 or 20 investments (i.e. invest in the top 6 or 12 performers)? Or would take the best performer of each asset class and compare that. In other words, down select from all domestic equity to get the best performer, and compare that one with the one best performer in bonds so you’re still investing in only three of five possible asset classes.

Lastly, any difference in changing how markets are divided in the broader portfolios? For example, does it make a difference if you divide the domestic market by sectors instead of by cap? Or how about dividing the overseas markets regionally instead of just emerging and developed markets?

For each I’m definitely not expecting fully reviewed, out of sample tests, etc. but would just like to hear your thoughts and input. You’re certainly generous with your time and any insights you have would be appreciated!

In your post titled "What a Difference a Year Makes! Endowment, Buy and Hold, and Tactical Returns" an update is provided for the Rotation Top 3. I would like to see an update the Rotation Top 2 and Rotation Top 1. thanks

I second an interest in how returns would vary on a different trigger date, specifically around the turn of the month.
I am curious how you would approach having a hedge fund (or strategy) as one of the assets classes. Not going short an asset, or using leverage, but an allocation to a strategy that is both long and short at the same time. Anyone comfortable with hedge funds likely has some simple allocation to this in their portfolio (HSGFX, JLPSX, or many others etc).

Over the past two years it seems that the asset classes have been moving together and becoming more correlated. I am quite curious as to
1) The what extent this increased correlation is really happening
2) Whether it is a temporary phenomenon or represents a shift in the market, i.e. more money being asset allocated, the increased use of leverage to invest in all assets, etc.

I was reading through some of these for ideas I'm sure you're familiar with.
Currency is the lead catalyst at present. Gold/PM's should be only slight mention, and personally I think the 5 asset classes covers it quite well.

http://papers.ssrn.com/sol3/papers.cfm?abstract...

http://papers.ssrn.com/sol3/papers.cfm?abstract...

How about some ways to improve on the Timing model. What other filters can be used beyond the moving average?

One of the things that holds me back is that, living in Canada, most ETFs available for International or U.S. eqity and bonds are in U.S$. The recent rise in the $SPX since March (and buy signal in July with the 10m-SMA) has resulted in relatively little gains due to the fall in the US$.

So, some discussion would be great on whether we should consider currency risk as part of the calculations (i.e. calculate 10month on $SPX:$CDW cross) or whether a better strategy is to hedge the underlying currencies.

Thanks for putting a great site and book together. I've found them to be very helpful.

The current research from SMAs presumes going to cash after falling below the 10-month SMA and become invested once it crosses above. I would be interested to see performance differences for going to cash after falling below the 10-month, but instead of waiting for the cross-over start dollar cost averaging back in (at pre-determined levels) as the market moves lower.

For example:

Falls below 10-month: 0% invested
Market falls 10% below original exit price: 10% invested
Market falls 20% bleow original exit price: 20% invested
Market falls 30% below original exit price: 30% invested
Market falls 40% below original exit price: 40% invested
Market falls 50% below original exit price: 50% invested

However, only once it crosses back above the 10-month is it 100% invested again. This could be altered to be more conservative/aggressive, but wonder how it would affect performance.

All this and more asset classes:
Small Cap Value, Small Cap Growth, Micro Cap:
International Developed, Emerging Market, Junk bonds, Timber

Moving average is an excellent indicator of momentum but it's not the only one. An obvious alternative is whether the X-day moving average rising or falling, so you may consider testing that. Also, you may consider combining more than one indicator (e.g. index has to be above the MA and MA has to be rising for a buy signal). I have done quite a bit of testing in excel and found rising/falling to be equally good and combining it with above/below to add incremental value vs. either one alone.

Random Doc,
I'd like to learn more about how you test on Excel. I've used Tradestation in the past but no longer subscribe. Thanks

A whipsaw reduction mechanism. I have used two in my testing with Excel that seemed to help. One is requiring two or more consecutive signals before triggering a switch. The other is requiring a buffer above or below the moving average before trigger a switch (e.g. index has to be X% above the MA before triggering a buy signal).

Something I have done in my own low-level Excel testing to reduce whipsaws was to use three monthly moving averages - 10EMA, 10SMA, & 12SMA. Only when the price is above all three would a buy signal be triggered and vice versa to sell. So while I know whipsaws are reduced with this method, I'm just not smart enough to backtest what this does for max drawdown, standard deviation, or total return over the long term.

A sensitivity analysis to test the robustness / brittleness of the 200-Day MA versus shorter or longer periods. Ideally you'd want to see a "mound of toast", where the absolute and risk-adjusted results increase smoothly to a peak close to your choice (e.g. 200-Day).

How about the studying the impact of rebalancing when an asset class' "share of total portfolio" increases by a certain percentages? (for example, rebalancing kicks in upon a 10/15/20/25 percent increase in "share of total portfolio" of any asset class).

1) Can you advise an easy to use software program that would allow us to actualize your model?
2) What about changing the weightings of the highest ranked classes (go 50/30/20 instead of 33/33/33 when buy the top 3 of 5 classes)
I too think highly of your ideas (both rotation and SMA) and would adopt one of your models immediately -- but everywhere I turn, gathering and processing the information is kind of laborious. Just looking for an easier way!

Hi. Could you suggest etf's that would substitute for the indicies that you examine in a new paper? If you decide to do so, could you either 1) include the correlation between the index and the etf or 2) run you analyses using the etf's instead of the indicies?

an appendix of the dates, prices etc of the trades.

A precise description of the best stop/loss strategy, particularly regarding what to do AFTER you've been taken out of a position. Should you leave it alone until the next month when you re-balance or should you do something else? And of course, what kind of stop/loss strategy should be used - a trailing stop/loss based on a percentage or what, and what would be the percentage?

Thanks...

I agree with a stop loss strategy test. Although the research is split on the value of stop loss strategies, I feel even if it add zero value on an absolute and risk-adjusted basis, it offer peace of mind knowing the maximum possible loss. This would be excellent protection against the black swan if you agree with Taleb's thesis. Also if you read the Market Wizards books, limiting losses is about the one thing every traded interviewed in the book agreed on.

One idea is to look at what day of the month is the best to rebalance. I suspect it would be a few days before the end of the month. The analysis might be a little tricky, since you would need to change data sources to get daily data and possibly look only at a more recent time period.

I've heard this before, but for the life of me can't understand why it would possiblly make a difference

I was a skeptic about the turn of the month effect until I ran some numbers myself. There really is something persistent there, in my opinion (in a test of 100 years with 20+ years out of sample in several different markets). I'd guess that the best time to get into new equities is a few days before the end of the month, an the best time to get out is a few days after. For some references online, see:
http://www.cxoadvisory.com/blog/external/blog7-...
http://www.cxoadvisory.com/blog/internal/blog2-...
http://papers.ssrn.com/sol3/papers.cfm?abstract...

From http://papers.ssrn.com/sol3/papers.cfm?abstract...
"In the attached study, we extend and broaden the Ariel and the L&S analyses using CRSP daily market returns for the period 1926-2006. To begin, we find that the effect occurs over the two decades that begin after the periods studied by them. For example, with VW returns, over the period 1987-2005, the mean daily market return over the four-day turn-of-the-month is 0.15%; whereas over the other 16 trading days, it is -0.001%. When our results are combined with those of L&S, on average, investors receive no reward for bearing market risk except at the turn-of-the-month over the 109-years of 1897-2005.

"We find that the effect is not confined to small-cap or low-price stocks; it is not concentrated at calendar year-ends or quarter-ends; and it is not confined to the US: of the 34 countries we study, the effect occurs in 30 of them. Further the effect is not due to risk as measured by standard deviation of returns. Finally, we find that the effect is not due to increased buying pressure measured by trading volume and net flows to mutual funds."

I have not analyzed turn of the month with commodities or bonds. Also note that the length of the turn of the month period has widened in the last few years (you can get in a day or two earlier now than in the 1980x-1990s, and still do well).

Meb,
most important on my list, so I'll cast my vote- the Global
rotation system
I also agree with Stephen - subcomponents of domestic stocks and Intl stocks (emerging mrkts)

Ok, this seems pretty popular so will probably add

One more suggestion for your January paper:

1. Explore impact of increasing the number of asset classes upwards from 5. Interested to know how adding further asset classes (or maybe dividing the existing 5 into subcomponents) impacts the return / volatility of your strategy.

FYI - bought your book the other day. It is a well explained and easy read. I am seriously thinking about applying the strategy as you define it in your paper to my 401K account

We have done posts about getting more granular within asset classes, not a
bad idea. What asset classes would you consider adding? (Currencies is an
obvious one.)

Yes, currencies definitely. In addition you may consider international bonds and commodities (energy, metals, grains, etc),

1. Yeah we have some posts on blog that address chopping up asset classes. What other asset classes would you add? Currencies are an obvious choice...

What I would like to see are tests of the stability of the coefficients. We all know that the 200-day mov avg works - that's why we look at it. But interest in these things has increased a lot. It may be about to be gamed out of the system. You should reformulate it as a arma process and see if it has been stable. If the coeffs have been declining or getting less predictive over time, it may be about to lose its value.

Well, seeing as how the model has the best performance ever out of sample in
real time, that would be a good test already, no?

Well, the out of sample results in real time were far better than the backtested results...that is usually the sign of a robust system

Yes that is a good sign. Still, I would like to see a quantification of it.