Following up on last night’s message, the “structural change” that “may” have occurred relates to the following thought: Virtually all of the research that I’ve done has been with market data from the end of the Korean War to the present. That’s currently a period of almost half a century. I had often considered going back further but have not for several reasons: (1) Pre-war, the percentage of the public that was involved in the market was much smaller than it is today, and the evaluation of market players from that era would have to be based on a rarefied strata of society compared to the broad groups of people investing today, (2) the market today is highly institutionalized with individual investing comprising a relative small percentage of total capital being traded; pre-war, the public (albeit only a small part of the public) constituted the vast majority of all market participation, and institutional participation was minimal, (3) the advantage gained by using a lot of antiquated information from a manufacturing economy in a Great Depression seemed somewhat without purpose inasmuch as a return to the economic circumstances of the Depression is highly unlikely, as is a return to a manufacturing, rather than a service, economy, (4) data from that era is severely limited, and some of it is of questionable accuracy.
However, the recent difficulties of my short term and daily models in accurately predicting the market led me to consider the following possibility: What if the predictions have been incorrect because we are in the early stages of a major Depression – not one with 23% unemployment or apple vendors or Hoovertowns – but a Depression of long term economic and psychological malaise, similar to what’s been happening in Japan over the past dozen years. In such an environment, unlike anything that’s happened in the US in the post-War era, how reliable as a predictor of the market would data from only the post-War period be?
The obvious answer is that it’s impossible to know until well in the future. However, there are ways of doing research on the subject, consistent with past research that I’ve done, that may provide useful answers, and that’s what I’ve been doing almost continuously since Sunday night. My preliminary conclusion is that while I intend to do more research on the question, it appears that the hypothesis just described might not be the answer. Rather, the fact that this is the first time in history that the Federal Reserve has aggressively lowered interest rates during the very, very early stages of economic contraction (such as I described in the February 5, 2001 issue of Turov on Timing) may be the most significant factor in identifying the culprit.
Interest rates and interest rate trends and the specific action/intentions of the Fed (even when such action is not immediately transmitted into changes in all interest rates) are a critical part of all my models, with more weighting applying to the longer term than to the shorter term, but with significance to all periods. However, since all aggressive Fed has occurred during late economic contraction periods rather than near the beginning of economic contractions, then the current environment is not reflected in any of the data that I use – or even any of the data that I conceivably could use — even if I went back to the beginning of the 20th century!!!!!
The fastest way (albeit not the most scientifically comprehensive way) to test for the significance of this second hypothesis is to run simulations that do not include interest rates or interest rate trends. Preliminarily, the results are startling – and consistent with this new hypothesis. Removing the interest rate inputs shows a market with all the characteristics of a massive bear market top and very little of the intermediate term bottoming that my models have been predicting.
So how do I reconcile this disparity? In the same way that I’ve always reconciled disparities, with further research and testing, followed by changes in the models based on the results of such research and testing. That will, of course, take some considerable time. In the interim, now that I see the significance of this factor, an adjustment in how I use the models, consistent with my preliminary findings – and in a manner consistent with other interim changes that I’ve made in the past while doing ongoing research – will occur. I will discuss this further later in this report to you.
Next subject: A very large part of Year 2000’s trading gains were a function of getting more and more aggressive by risking accumulated profits in an attempt to magnify them when they occurred, coupled with becoming more and more conservative when losses occurred. So, looking at the experience of October, 2000 is very worthwhile when evaluating the recent trading losses.
During the first four days of October 2000, equity in my clients’ accounts declined by almost 34% — a shocking loss in only four trading sessions. I responded by becoming very conservative and much more selective in the trades I took. They weren’t all profitable, but the majority were, and by the end of October, I had pared the losses in half. It still was a losing month though. In November, we made up the remainder of October’s losses and had a small profit compared to September’s closing equity.
I regard the current period as similar to October 2000 in that we’ve had a nasty drawdown and the question is how to deal with it. The answer is “the same as I did last October,” by “becoming very conservative and much more selective in the trades” I’ll take. Complicating matters, of course, has been the interest rate input factor that I discussed a moment ago – but now that I’ve identified it, working around it should make things much easier.
Next subject (but related to the previous one): No one, not I, not anyone, can predict the market. All I can do is (1) predict the probability, in a news-neutral environment, of the odds of the market advancing or declining in a given day, (2) predict the theoretic most likely magnitude of the advance if the market does advance, and the theoretic most likely magnitude of the decline if the market does decline, and (3) factor in the probability of the prediction being accurate (what I call the risk component of the daily model). So, for example, going long yesterday had a 55% probability of being right and the magnitude ratio was 1.4. So, somewhat simplistically, the opportunity to risk ratio of going long was 55 divided by 45, times 1.4, or 1.7 to 1. Over time, numbers like that will work out very, very well. In any given day, they still have a 45% chance of being wrong – as was the case yesterday when we went long and were stopped out with a $3700 per unit loss, leaving us with a net loss for July to date of $2950 per one normal trading unit account size. Of some considerable significance, re-running the numbers for yesterday with the temporary interest rate adjustment I’ve just instituted would have resulted in an opportunity to risk ratio of only 1.2 to 1 – a level below acceptable parameters during conservative times.
I intend to immediately run simultaneous sets of data on all short term, daily, and intraday models using both the tried-and-true existing parameters as well as the new untested interest rate adjusted parameters. In order to qualify for a trade, the opportunity to risk ratio must be at least 1.5 on BOTH sets of data. At some point in the near future, when my research on the interest rate issue is completed, I will be able to integrate this new factor into the existing models, with the result being an improved refined model. By the way, this type of refinement is not remotely new to my models. I have done this many times over the years in response to changing market circumstances, and the mechanism for my doing this is not experimental. Finally, although I consider my clients to be my friends, I’m sure that if my lawyer were looking over my shoulder he would encourage me to add the following statement: The fact that I’m discussing these issues with you does not in any way diminish the Page 4, paragraph 1 statement in my Disclosure Document that my “methodology is completely proprietary,” and all the related text that follows it there.
The adjustment described in the preceding paragraph should further moderate the number of trades that I take until we have the Net Asset Value of your account moving solidly upward. That’s the strategy I employed last October, and that’s the strategy that I will employ now. I’m just relieved that I finally can see WHY the model has been so stubbornly incorrect – and how it can be corrected. The simulation of yesterday’s signal – and the quantifiable reduction of the opportunity to risk ratio from 1.7 to 1.2 – is particularly significant to me, inasmuch as it is the first quantifiable adjustment that I’ve been able to make since the NAV slide began.
It’s 5:30 a.m., Pacific Time, and I’ve been working all night. I’m going to meditate and shower, and I’ll be back in the office at 6:00, a half hour before the market opens. Preliminarily, I don’t see any trading opportunities at present, and as I’ve clearly indicated, I’m not going to push on a string; I’ll wait for the double 1.5:1 opportunities to arise, and I’m cautiously optimistic about repeating the October/November experience from last year over the balance of the summer of this year.
Best wishes, Dan