Ask Colin

Some time ago you said you thought the various American authors of books on (the various variants of the "Dogs of the Dow" method) had just optimised their back testing to sell books, i.e.. Michael O' Higgins (Dow 5) and the Gardiner Bros. with their Dow 4 and Foolish 4. Are you still of this opinion? If so why?

I think you are misrepresenting what I said. The Dogs of the Dow" approach is to invest equal amounts in the 10 highest dividend yields from the 30 Dow Jones Industrial Average stocks and hold them 12 months, repeating the exercise annually. This has been tested over a long period and beats the market. This method was developed by James O'Shaughnessy.

Other writers have set about improving on it by taking smaller numbers of stocks from the 10 in "Dogs of the Dow". They have run tests on these smaller numbers of stocks and shown that they perform better than buying the whole 10 stocks.

What I tried to express to you was my concern that this approach was rather akin to the over optimising error common in trading systems. What this means is that you find something that works on a set of data. Then you keep tweaking parameters or adding additional parameters until you get the best possible result. The danger in this approach is that you are optimising for a specific data period.

System developers have ways of ensuring that they do not fall into this error. There is a good description of the process in Colby and Myers The Encyclopaedia of Technical Market Indicators part 1.

I have not read thoroughly the books that describe these methods (lack of time and higher priorities rather than lack of interest), so I can not judge the extent to which they may have fallen into this trap.

Your exploration of another data set - the Australian market - may go some way to helping us judge this issue.

There is one other concern that I have with the approaches that buy fewer than the 10 "dogs". This comes to the fundamental issue of risk. Reward should always be balanced by a consideration of the risks involved. It is well known that you can tweak reward higher by assuming greater risk.

One of the basic ways in which specific risk is managed in the stock market is diversification. The 10 "dogs" is already an aggressive strategy from the point of view of diversification - especially in Australia, where the "dogs" routinely come up with 3 or 4 banks in the 10 stocks. By concentrating on only 4 or 5 of the "dogs", it seems clear to me that the risk will be higher - expressed as greater volatility in returns. You seem to be confirming this in the comments you made on the question.

For this reason, I think it is important in comparing the results of the various methods that you include a risk calculation - say a standard deviation calculation alongside the returns achieved.