February 1997
Letters to the Editor
Editor,
BOOKMARKED
I am a novice trader, having been introduced to the wonderful (and often bewildering) world of trading futures and commodities.Today was my first visit to the S&C Web site - and I spent more than three hours absorbing all the information and the excellent links.
Keep up the good work. I have added you to my bookmarks and intend visiting as regularly as possible.
UDAY KILPADY
Sydney, Australia
via E-mail
Editor,
RELATIVE STRENGTH RATIO
In your interview with James O'Shaughnessy in the November 1996 issue, you refer to a high relative strength ratio. I know how to calculate a relative strength index (divide the stock price by the market index), but I've never heard of a relative strength ratio as used in the context of your question in the interview. How is it calculated?CARL SWENLINIn James O'Shaughnessy's book, What Works on Wall Street, the appendix lists:
via E-mail
Annual relative strength excludes dividends and uses simple price appreciation:
(PRCC/PRCC[-1y]) where: Prcc is the price of the stock today; and Prcc[-1y] is the price of the stock one year ago.-- Editor
Editor,
PSEUDORANDOM PRICE SERIES?
In reference to "The siren call of optimized trading systems" by Dennis Meyers in the October 1996 issue, Meyers reported some results from optimization performed on a randomly generated price series that he created. His conclusion in part was, "Clearly, a system can be optimized and produce good results even on a random series of prices."This is new and disturbing to me. Put simply, mechanical or visual optimization looks for patterns within a price series. This is often called curve-fitting. I wonder if his random price series was really a pseudorandom series. Let me expand on this concern.
A computerized random-number generator was used in generating his price series. In Programming Classics by Ian Oliver, an entire chapter is devoted to random-number generators. He begins by stating, "Doing things at random would seem to be very easy to accomplish. With a computer, it is actually quite difficult to do it well." He later states, "Random numbers generated by a computer are not really random, but are carefully calculated to look as if they have no plan, purpose or pattern." For this reason, they are called "pseudorandom." After reading the rest of the chapter, I would add, "as if they have no plan, purpose or pattern for the intended use."
Packaged computer pseudorandom-number generators may be acceptable for producing relatively short series of random numbers, but I wonder about their use in producing a relatively long series of numbers for testing, as was done in the article. In long series, patterns brought about by their pseudo nature have a chance to show up.
In view of these considerations, I question whether the price series used in the article was sufficiently random so as to not contain price pattern.
HUGH L. LOGAN
Hendersonville, TNDennis Meyers replies:
Mr. Logan is correct in his observation that computer-generated random numbers are only pseudorandom in the sense that computer-generated random series repeat after n random numbers and are dependent on what is called a starting seed. However, n is on the order of 10,000 to one million or more numbers, depending on the type of random-number generator used and the computer used.In my article, I generated only 1,000 random numbers using Microsoft Excel's random-number generator on a Pentium computer. This is well below the level where the number series could begin repeating.
On a related topic, readers may want to look at Introduction to Probability Theory and Its Applications, Vol. I, by William Feller, page 84. Pseudoprices created from random data quite frequently produce visually obvious trends. Trends are the norm in random processes, not the exception. When a system is applied to a random price series, zero net gains occur on average. This is because, on average, the whipsaw losses from a system wipe out all the system's profits gained by trending prices. However, in a short random price series, where the "on average" doesn't have time to kick in, a system can always be found by curve-fitting to produce noteworthy returns. A number of analysts have claimed that curve-fitting 20 years of daily data captures most all of the market price dynamics and out-of-sample testing is not needed. A daily price series of 20 years is about 5,000 prices. In random process theory, 5,000 is a small number of samples. I have shown in my article how easy it is to curve-fit 500 random prices. It's almost as easy to curve-fit 5,000 random prices. Thus, without careful out-of-sample testing procedures, it's impossible to determine whether a system has curve-fitted a random series or has captured the dynamics of a real price series.
Contributing Editor Dennis Meyers has a doctorate in applied mathematics in engineering. He is a member of the Chicago Board Options Exchange (Cboe), a financial institution consultant and a private trader. He can be reached through E-mail at meyersx@enteract.com. -- Editor
Editor,
POPULATION SHIFT AND THE STOCK MARKET
I agree that comparing US birthrates to the S&P 500 index, adjusted for inflation, is persuasive (Opening Position, December 1996 STOCKS & COMMODITIES). I am not certain that it is bullish. As of mid-November 1996, the S&P 500 stands at 737. The chart on page 6 of the December issue implies that in the year 2008, the S&P high will be 860. Most of the advance occurs over the next year or two and in total represents a small 17% movement.In 2008, when today's 50-year-olds are 62 and beginning to raise the level of cash in their 401(k) plans, we can anticipate a downturn of about 50%. In 20 years or so, if you've factored in inflation, the S&P 500 will be at a lower level. This is no reason to own stocks.
Peak momentum and rates of change in the stock market are surely behind us.
HAL CUMBERLAND
Wichita, KS
via E-mail
Editor,
PSYCHOLOGY AND THE UNIVERSALITY OF TECHNICAL ANALYSIS
This is my first letter since I began subscribing to your magazine more than three years ago. Maybe the ease in communicating via E-mail motivated me to write.Like many other S&C readers, I recognize the high educational value of STOCKS & COMMODITIES and I have to say that I have found in it most of my trading ideas, as well as references to many interesting authors and software developers. Maybe even more interesting, I've discovered the articles that didn't catch my personal interest during the early days I subscribed (articles such as those on trading psychology, habits and behavior) should have been the first I focused on!
One thing that continues to amaze me is the universality of technical analysis: If an indicator or a trading system really works, it will work in any market, from the United States to Europe, from Asia to Australia. This is probably the beauty of the law of supply and demand.
Maybe you could help me with a question:
I'm considering the purchase of a set of books (that are presented as a trading course) on the theory and application of market cycles. The problem is it's not cheap! Before purchasing any software or books on financials, I always try to check previous STOCKS & COMMODITIES issues for reviews, and I've often found them for the products in question. Except in this case. Could you help me? I know you always try to stay very neutral and objective, so I'm not asking for a rough opinion like "bad" or "excellent" but only a bit more information, if you could. The work seems derived from Gann work.
Thank you, and again, congratulations on your work. Please keep on.
CARL VANHAESENDONCK
Brussels, Belgium
via E-mailSorry, we have no information on the book set you mention in your letter. We suggest you request information from the book's publisher or research the book on the Internet by checking the sites of book distributors. -- Editor
Editor,
PACKED WITH INFORMATION
Bravo for another great article by John Ehlers in the November 1996 STOCKS & COMMODITIES. Ehlers has shared many great concepts in the past in S&C. Writing in the highly condensed mathematical format that he does packs a lot of information into a few pages. The ideas he presents always seem to work when I experiment with them myself. (That's probably why his MESA software package has been so successful.) Any of the key concepts he mentions, such as dominant cycle, the sine wave indicator, and so on, could serve as the basis for additional articles or for Traders' Tips contributed by S&C staff or various product vendors. (By the way, could you forward a copy of your contributor guidelines to me? I may wish to follow up on my own suggestion!)The article, however, did bring up a question that I'd like to direct to Ehlers:
In this and prior articles, he discusses the dominant cycle, which can be identified by various mathematical techniques, such as those employed in his software program, MESA. However, it seems that the critical cycle of interest to a trader is also relative to the trader's style and preferred time frame -- which are not necessarily known by the mathematical model. Therefore, the selection of the dominant cycle by the mathematical techniques employed must contain underlying assumptions about these factors. Could Ehlers briefly explain the assumptions he is using? Perhaps he could provide a working definition of dominant cycle.
In past articles, he's described conceptually how his mathematical model captures the appropriate characteristics of amplitude and other cyclical components to identify the dominant cycle. In "Early trend identification" (October 1990 S&C), Ehlers describes the power spectrum that is characteristic of the dominant cycle. He used this concept in examples in his later articles. But how does that fit with the trader's requirements? In short, I'm asking why he believes it works rather than how it works.
On another note, let me also applaud the series of trading psychology articles by Adrienne Toghraie in S&C. They have been excellent!
ARTHUR CHESTER
Rochester, NY
via E-mailJohn Ehlers replies:
I am always delighted when readers find value in my work. It is my goal to bring science to the art of trading.The analysis span used to find the dominant cycle is mostly a factor of the dynamic range of the measurement technique. The MESA measurement is valid for the eight-to-one range from six to 50 bars without adjustment of internal parameters. Since we are using sampled data, it matters very little whether those samples are daily, weekly or hourly bars. In The Profit Magic of Stock Transaction Timing, J.D. Hurst established the proportionality principle -- that is, the longer the cycle, the bigger the amplitude. Thus, one can trade cycles over any period he is comfortable with and expect the performance per trade to be proportional to the length of the cycle.
I don't only believe cycles work, I know they work. Cyclic behavior is ubiquitous. The natural response to any physical disturbance is harmonic motion. For example, if you strum a rubber band stretched between your fingers or bend a ruler over the edge of a table and release it, you create an oscillation. This cyclic reaction to disturbance holds from the atomic level to colliding galaxies. Since there are plenty of disturbances in the market, there is no reason to think that harmonic motion should not apply here also. The philosophical underpinning of my approach lies in two solutions to "the drunkard's walk" problem. The two solutions are partial differential equations called the diffusion equation and the telegraphers' equation. I develop these in greater detail in my book, MESA and Trading Market Cycles. By relating the market to physical phenomena, I conclude these equations are the general descriptions of the trend mode and cycle mode. It's crucial to know the mode of the market to establish the correct entry and exit strategies. The cycles themselves come and go and drift from one length to another. It's important to pinpoint this activity as it's happening.
Editor,
REGISTRATION OF COMMODITIES FUNDS
Thank you for publishing STOCKS & COMMODITIES, a truly informative trade journal. I am an independent trader and wasWhat avenue, in your opinion, should I use to register my futures fund? I would like to be able to establish a track record vis-à-vis other futures funds. Is there a low-cost way to get registered? The fund is worth about $300,000 and has only one investor -- me.
MARK LARKIN
via E-mailI can't provide information on fund registration. I suggest you contact the Managed Futures Association at 415 325-4500, E-mail mfa@mfainfo.org, or www.mfahome.com for information. -- Editor
Editor,
CHAT FORUMS FOR TECHNICAL ANALYSIS
Can you provide a list of chat rooms and or other Web sites where information pertaining to technical analysis will be available and/or discussed?I enjoy your magazine very much indeed. Lots of things don't make sense to me yet, but someday they will and I keep on reading and trying to get somewhere. So many stories appear in the news about financial analysis; where would I have been had I not been a STOCKS & COMMODITIES avid reader?
ERNEST O. JENSEN
Seattle, WAOne Web site address I can list is www.realtraders.com, which is not a chat room but a discussion group sponsored by Kasanjian Research. In the future, we hope to add discussion forums to our own Web site at www.traders.com. -- Editor
CONGRATULATIONS
S&C January 1995 art by Carolina Arentsen We'd like to congratulate STOCKS & COMMODITIES illustrator CAROLINA ARENTSEN of Hint Studio for winning an award in Print magazine's regional design annual competition. The winning illustration (shown above) originally appeared in the January 1995 STOCKS & COMMODITIES.
