Daily Speculations The Web Site of Victor Niederhoffer & Laurel Kenner Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter;  a forum for us to use our meager abilities to make the world of specinvestments a better place.

Home

Write to us at: (address is not clickable)

The Sun-Baked Speculator
Tom Ryan

09/29/04
Geology and Markets

For years miners, especially underground miners talked about earth tides in the rock in response to the moon. And there have even been some countable studies, namely in large caving operations where draw data would suggest the existence of response in the mine to the moon cycle. But of course there are so many overlapping influences and much noise and precision errors with regard to measurements. But recently, as technology has improved, particularly Doppler and ground penetrating radar, there is increasing evidence of ground response in large rock structures to the moon. It's very small, almost unmeasurable, but it's probably there.

For example, although the sample size is small, we have documented a tendency for very large rock slides in open pit mines to develop during full moon periods. Of course, this small sample size presents a problem for when it comes to predicting geologic events, and this Parkfield business and the eq swarm at St. Helens are good examples, we have the classic problem of the base rate vs the case rate. When it comes to very long term records, say 500 years or longer, we don't have a lot of great records but where we do have it, say floods on the Nile, Vesuvius, geomorphology of fault scarps, volcanic ash layers in soils, tree rings, to name a few, in nearly all cases we see negative exponential or Weibull distributions. However, data from the most recent 100-200 years can take almost any form. This leads to the dilemma, do we use the actual case rate as measured, and what if that case rate does not take the exponential form as we tend to see with events that have the longer records.

The market analogy is clear, how far back in the record do we test our systems. And what does it mean to see a system work for one period and not in the next. We have all probably had that experience of developing a trading system only to be confronted upon implementation with three or more losers in a row.... what does it mean to have a back-tested system that suddenly fails 3,4, times in a row? (up until today the many down afternoons lately come to mind). It's the same base rate vs case rate problem. Of course, there is a big difference between markets and rocks, people learn and adapt. The changing cycles ala Bacon. But the length of the back-test is always problematic it seems to me. Henry and i talked a bit about the application of GA's to market trading systems and one area which I have toyed with is using GA's to determine fitness verses length of back-test because GA's are such good optimizers of noisy data. But my initial work on it just generated another noisy time series as fitness seemed to jump back and forward in time. That was just one system however so maybe applying it to another would yield a substantial meal.

This is a long winded way of saying I have no clue when St. Helens will erupt in a big way again, although the precursors like an eq swarm (or rising volatility in the case of prices) might be an indicator of something on the horizon. Given the prevailing wind direction at this time of year there would be some interesting speculative implications.