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: mauser96 who wrote (25416)	5/27/2000 1:17:00 PM
From: Thomas Mercer-Hursh	Read Replies (1) \| Respond to of 54805

return per unit of risk or standard deviation Apropos of which, I played some interesting games this week with the long term history of weekly closes for the S&P500 and the NASDAQ. I should note up front that while I am a sometimes mathematician, I have never had any interest in TA because mathematics to me has always been in service of some biological or social problem which I was investigating. E.g., if the issue was exploring the influence of size on shape in a population of primates, the mathematics was only useful or interesting if I could tie back its results to something I could interpret in biological terms. Without that, the math is merely descriptive. When TA types start talking more about how the math provides some insight into underlying forces, or, better yet, when the math starts coming from a theory, I might get more interested. Anyway, I have a naive friend who was doing some retirement planning on an overly simplistic model and I wanted to impress her with the concept that such and such an average annual gain wasn't at all the same thing as what one would actually experience invested in the market. So, I downloaded a 50 year history of the S&P and '84 to date on the NAZ. For each I calculated a trendline based on a 52 week moving average window (not the trailing average one usually sees, but +/-26 weeks since that follows the trend better). I then calculated the absolute and percentage deviations from that trendline and plotted the curves for the actual, trend, and +/- 3 standard deviations from the trend, i.e., the range that should enclose 99% of the observations. Well, the +/-3SD lines did enclose most of the variation and, not too surprisingly the variation of the NAZ was significantly higher (1SD=7.06%) than the S&P (1SD=3.95%). But, one of the things I found interesting were the places where the curves violated the +/-3SD envelope. For the S&P, one of the very few of these was late summer 98. The recovery from that was strong, but didn't overshoot. With the Naz, however, last fall there was a dramatic violation on the high side, enough that as a statistician I would be sent scrurrying to check my measurements. At peak, it was 4.86 SD above trend, almost to that vaulted 5 sigma realm. And the counter reaction has now gone 3 sigma on the negative side in quite a short period. This is still descriptive statistics, of course. There is no model here as to why these variations occurred. What it tells us, though, is that without understanding why, there clearly has been a pattern of growth will outside any normative historical range ... something one might well be tempted to suggest indicated systematic overvaluation, followed by a correction, an overcorrection, perhaps due to the same kind of phenomenon one sees in any system with negative feedback controls when observed values exceed the range the controls were designed to maintain. FWIW, similar pattern occurred in the S&P in late summer 87 when it went to 4.52 positive SD the week of 17 August and then corrected to a 4.81 negative SD the week of 30 November. One of the things that this little game reinforces, by the way, is the importance of judging things in relationship to their current context. Looking at absolute metrics will give you a very different picture than relative metrics, i.e., point changes versus percentage changes.