Message Font: Serif | Sans-Serif
No. of Recommendations: 44
I'm writing the following message to warn you all about two of issues. I will not be concerned with the data mining issue but merely discuss all available out-of-sample evidence, that is how the strategies work in practice, then it's up to you to judge the reason for the poor out-of-sample performance of the most important DDA strategies.

First issue: Returns haven't been spectacular lately

For example, the following was stated in an official FF column:

Returns haven't been spectacular for the last several years

At another board a TMF'er stated:

one or two years of under-performance don't mean anything

Let's look at the numbers for a bunch of important DDA approaches.

HY10 &HY5

These are the original DDA strategies. HY10 for sure and possibly the HY5 were developed by Slatter together with Knowles and Petty using 1972-1987 annual data.


In the first version of his book O'Higgins developed the BTD5 using 1973-1989 data.


The first official TMF Foolish Four. It was developed using 1973-1993 data.


Current official Foolish Four developed by TMFElan in 1997and TMFSandy in 1996 (before they became
TMF'ers). That is, the out-of-sample period could be defined as 1997-1999.

YEAR HY10 HY5 BTD5 F4.0 F4.2 DOW30

1988 17.96% 20.58% 13.78%
1989 29.64% 28.54% 31.95%
1990 -10.01% -6.95% -17.34% -9.14%
1991 35.24% 64.32% 59.14% 30.36%
1992 6.35% -4.76% 19.95% 11.00%
1993 23.54% 21.58% 30.49% 17.91%
1994 2.43% -6.05% 6.26% -6.71% 3.73%
1995 37.10% 42.27% 30.13% 37.49% 36.69%
1996 27.47% 26.95% 26.62% 30.47% 24.32%
1997 20.39% 21.47% 17.70% 24.30% 19.49% 22.33%
1998 9.67% 12.28% 12.49% 17.35% 15.64% 15.99%
1999 3.39% 7.77% -4.60% -14.00% 24.22% 22.28%

CAGR 16.06% 17.38% 16.37% 13.16% 19.73% 17.77%
DOW30 17.77% 17.77% 16.84% 20.48% 20.16%

The above is based on a table I wrote to the MI board, except that the 1999 numbers are slightly different. For some reason I had a stock in my DDA's that TMF didn't, so results I have shown previously are slightly off. However, it doesn't change anything. All of the above numbers are now from TMF sources:

Note that, while the official F4.2 underperformed the DOW30 in 1999, the above numbers indicate outperformance instead. This problem arises because the above numbers are based on 12/31/98-12/31/99 returns, while the official F4.2 now uses 12/24/98-12/27/99.

The second row from below displays CAGR's for the different strategies during their entire life. That is, the period after the historical data that was used in developing them. The last row displays the CAGR for the DOW30 during the same period as the respective strategies, i.e. the numbers to beat.

The table indicates that all of the important DDA strategies considered have average returns (more precisely CAGR's) lower than the DOW30. That is, during their entire life no strategy has outperformed the DOW30!

During the first out-of-sample year (1988) the HY10 and HY5 outperformed the DOW30, the second year they underperformed, the third one outperformed, the other underperformed. In the end they both lost, although clearly not statistically significantly. Therefore, the strategies have never worked in practice during their twelve years of out-of-sample performance. The BTD5 underperformed the first year, then hammered the DOW30 in the second year. However, its subsequent underperformance has been sufficient to give the strategy an out-of-sample CAGR similar to the DOW30. Therefore, the BDT5 has never worked.

The F4.0, i.e. the original FF, has grossly underperformed the DOW30. However, this is mainly due to 1999 where it was crushed by the DOW30 by 36%. Eliminating the 1999 return, it has a CAGR that is virtually identical to that of the DOW30. Therefore, the F4.0 has never worked in practice. Finally, the current F4.2 and the DOW30 have performed similarly during the three years of out-of-sample data.

As you can clearly see from the above none of the important DDA strategies have ever worked after the data they were developed on! Furthermore, it's not just the last one, two, or several years. The lack of out-performance is universal and applies to all out-of-sample data.

If somebody in the future claims that TMF is moving on to a new strategy because the old strategies have stopped working, it's simply not true. The strategies never worked!

Also, I didn't want to get into a discussion about Sharpe ratios, i.e. a way to risk-adjust the return. The description of Sharpe ratios in the TMF DOW spreadsheet is misleading, and I don't want this to turn into a new discussion about the Sharpe ratio. However, it's worth mentioning that the standard deviations of the 4-5 stock DDA portfolios are 50%-100% higher than that of the DOW30 out-of-sample. Therefore, all considered DDA strategies would grossly underperform the DOW30 in risk adjusted terms.

Second issue: DOW stocks have lost their attractiveness, so we need to move on to a different/ expanded set of stocks.

For example, the official FF column stated:

Of course, we like the Foolish Four best because it has a long-term history of beating the market. Right now, though, the Foolish Four is a strategy in transition. Recent changes to the Dow have altered the nature of that list somewhat, reducing the importance of strong dividends as criteria for being a Dow stock. Although it has only become obvious recently, this trend has been going on since 1991, and at this point, I'm not so sure that the Dow is the best pond for us to be fishing in. A list of major blue-chip stocks that pay significant dividends, whether they are on the Dow or not, might provide more consistent returns. We are checking out that possibility.

I assume they have the S&P 500 stocks in mind to get big issues. TMF is also following the so-called BSP strategy, which as far as I understand is close to a F4.0 implemented on 30 big S&P 500 stocks that are not in the DJIA, and are not utility companies. Here is some data from official TMF sources:

Mix in- and out Out-of-sample?

1987 5.23% 13.10%
1988 16.81% 32.10%
1989 31.49% 25.80%
1990 -3.17% -4.30%
1991 30.55% 25.50%
1992 7.67% 3.90%
1993 9.99% 20.20%
1994 1.31% 5.60% 1.31% 5.60%
1995 37.43% 40.00% 37.43% 40.00%
1996 23.07% 33.50% 23.07% 33.50%
1997 33.36% 33.00% 33.36% 33.00%
1998 28.60% 28.90% 28.60% 28.90%
1999 19.50% 2.30% 19.50% 2.30%

CAGR 17.89% 19.15% 23.29% 22.99%

This is all the data we have. The first two columns labeled "S&P500 " and "BSP" give the annual return since 1987 thus mixing in- and out-of-sample data. This is the data TMF uses in promoting the strategy. As you can see there is not much more than a percentage point different in returns. Thus, the BSP hardly beats the S&P 500 even when mixing in in-sample data. Using the same out-of-sample period as for the F4.0, the BSP clearly doesn't beat the S&P 500. Thus, there is no evidence that DDA's work on other sets of stocks. Notice that the BSP ought to be compared to the set of 30 S&P 500 stocks it was selected from, and not the S&P 500. Also, the BSP is about 25% riskier than the S&P500, so it substantially underperforms in risk-adjusted terms.

It doesn't help to fish in a different pond if you are using the wrong bait!

In conclusion, the DDA strategies simply don't work in practice. They never worked. They never worked on a different set of stocks. There is no reason to believe that any future refinement will ever work, not on any data.

The author formerly known as Datasnooper.
Print the post  


When Life Gives You Lemons
We all have had hardships and made poor decisions. The important thing is how we respond and grow. Read the story of a Fool who started from nothing, and looks to gain everything.
Contact Us
Contact Customer Service and other Fool departments here.
Work for Fools?
Winner of the Washingtonian great places to work, and Glassdoor #1 Company to Work For 2015! Have access to all of TMF's online and email products for FREE, and be paid for your contributions to TMF! Click the link and start your Fool career.