Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Gorilla and King Portfolio Candidates -- Ignore unavailable to you. Want to Upgrade?

To: Rick who wrote (32697)	10/1/2000 3:55:58 PM
From: mr.mark	Respond to of 54805

"Why would the discarded strategies have anything to do with the successful one?" because in an environment where all 100 tests are not conducted simultaneously, one strategy may have succeeded merely due to market timing, and the others may have failed for the same reason, albeit at a different time, thus giving a false read. no? :) mark

To: Rick who wrote (32697)	10/1/2000 9:54:53 PM
From: Mike Buckley	Respond to of 54805

Fred, If Mark Hulbert doesn't think there's any basis for choosing the lowest priced of the highest-yielding Dow stocks, he needs to look no further than Micahel O'Higgins's book, Beating the Dow to become educated on the subject. O'Higgins did the individual investor a huge favor by doing the research and writing the book. Ann Coleman of the Motley Fool was quoted at the end of the article. Hulbert responded that her approach sounds like data mining to him. Since Mark's office is only blocks from Ann's office, he would do well to spend more time with her. Had it not been for Ann Coleman and her painstaking research about six years ago that applied common sense and fundamental analysis -- not data mining -- to mechanical investing, I'm not sure I would ever have come to appreciate the value of the online community. She and the people who worked with her opened my eyes about the credible opportunities the online world brings to each of us. At the time Ann was not a volunteer nor an employee of the Fool. She was exactly like each of us here in this folder, hoping to seek truth about the past. When I achieved the financial independence that allowed me to quit working, Ann was on a very short list of people I personally thanked for giving me the tools needed to attain such good fortune. I think it's especially ill-informed of Hulbert if he doesn't realize (at least he didn't write about it) that the Beating-the-Dow method of mechanical investing has a very good chance of repeatedly underperforming the Dow. Why? Because Microsfoft and Intel are now in the DJIA. Neither of them pay dividends. As a result, neither of those two stocks will be selected by that mechanical method yet they have a reasonable shot at outperforming almost all the other components of the DJIA if not all of them. Like any investing methodology, it needs to be scrutinized. But calling Ann's work something that it isn't -- data mining -- is irresponsible journalism, or whatever it is that Hulbert calls his writing. --Mike Buckley

To: Rick who wrote (32697)	10/2/2000 3:58:53 AM
From: tekboy	Respond to of 54805

I actually did agree with it, and was thinking of posting it also. Key distinction for me is between stumbling across superficial historical patterns via data-mining and thinking theoretically, that is, analyzing the causal mechanisms that will generate real patterns both past and future. Gorilla Game does the latter. tekboy/Ares@exsocialscientist.com

To: Rick who wrote (32697)	10/2/2000 11:17:01 AM
From: tekboy	Read Replies (1) \| Respond to of 54805

Why would the discarded strategies have anything to do with the successful one? I think I came up with the answer to this one. If the "strategy" in question involves betting on the recurrence of a pattern that appears in historical data, you better be sure the pattern is "significant" and thus is likely to persist in the future. One way of ascertaining this is to locate the causal mechanism for the pattern's appearance and determine that it is still going strong. A second, less intellectually challenging way is to use tests for "statistical significance," i.e. what the odds are that the pattern might have appeared by chance. It's in that second case that the number of strategies tried and discarded becomes relevant: If I test a thousand "strategies" on a sample of historical data, then by definition it's highly likely that a few of them will post amazing results that seem like one-in-a-thousand shots. If I then throw out the rest and say to you "hey, try this one, look how well it did," you are being primed for a sucker bet because there is no real reason to believe the strategy will continue to outperform. tekboy/Ares@thinkthat'sright;remember,I'minnumerate.com