August 2004 Letters To The Editoror return to August 2004 Contents The editors of S&C invite readers to submit their opinions and information on subjects relating to technical analysis and this magazine. This column is our means of communication with our readers. Is there something you would like to know more (or less) about? Tell us about it. Without a source of new ideas and subjects coming from our readers, this magazine would not exist.
Address your correspondence to: Editor, STOCKS & COMMODITIES, 4757 California Ave. SW, Seattle, WA 98116-4499, or E-mail to editor@traders.com. All letters become the property of Technical Analysis, Inc. Letter-writers must include their full name and address for verification. Letters may be edited for length or clarity. The opinions expressed in this column do not necessarily represent those of the magazine. -Editor
ZIGZAG INDICATOR VS. FILTERED WAVES
Editor,
As one who has more gray in his beard than he'd like, I wanted to point out that the "zigzag indicator" or "zigzag filter" referred to in Norman J. Brown's June 2004 article is misnamed. I believe the mistake started when Equis included this indicator in the first version of its MetaStock charting software. CompuTrac did not include the indicator in its package (I told you there was too much gray in my beard).
At any rate, the concept was introduced before IBM had even introduced the first PC. It was originally introduced by Arthur Merrill in his book Filtered Waves: Basic Theory. That's the proper name for the indicator -- Merrill's filtered waves, not "zigzag." That book is filled with an amazing amount of pre-computer calculations and insights. Merrill didn't start in technical analysis until he retired from GE, after what most would have considered a full career. The Market Technicians Association honored him with their annual award years ago for what he did in his second career as a technical analyst. He is still alive and well (in his 90s, I believe).
Just as we no longer call the Arms index by the old Quotron symbol for it (TRIN), we should use the term Merrill's filtered waves, or at least use Merrill's term, filtered waves, when referring to the indicator. There are few truly new ideas in the world, much less so in this business. When they are introduced, we should give credit where it is due.
John Carder, CMT
Topline Investment Graphics
www.topline-charts.com
or www.chartguy.comThank you for pointing out that what is commonly referred to as zigzag can be the same as or based on the concept that Arthur Merrill originally introduced as filtered waves.
Readers can find our interview with Arthur Merrill, "First Citizen Of Technical Analysis: Arthur Merrill," in the October 1992 S&C (our 10-year anniversary issue). We published "Merrill MW Waves" by Merrill -- who is a chartered market technician (CMT) with the Market Technicians Association -- in our November 1991 issue. In addition, Merrill contributed more than 50 articles to this magazine in earlier years, and in recent years he has written letters to the editor that have appeared in this column. Anyone interested in Merrill's articles can use the search engine at our website, www.Traders.com, to locate and purchase them.
Some of Merrill's books are available through Windsor Books at www.windsorpublishing.com, from Equis International at www.equis.com, and other sources.
Readers can learn more about the Market Technicians Association at its website, www.Mta.org.--Editor
DURBIN WATSON STATISTIC
Editor,
I am writing in response to Ron McEwan's April 2004 article, "Come Here Quick, Durbin Watson." The author cites the use of the Durbin Watson statistic as a means to test for autocorrelation and that it is better than using normal correlation. He says that using autocorrelation is very useful in implementing a pairs strategy.
I believe that the author is incorrect in this assertion. The presence of autocorrelation is more an indication that the regression model being implemented is misspecified. In other words, the linear regression model is not using the correct variables (or combination of variables). When an equation shows a high degree of autocorrelation, the result is not an opportunity to benefit from pairs trading, it is an indication that the model needs to be re-evaluated. This is because the coefficients from the linear regression will be misstated when there is autocorrelation, since one of the major assumptions in OLS (ordinary least-squares regression) is that the errors are distributed normally with a mean of zero. Hence, using deviation from the regression trendline will be inaccurate, since the regression equation is wrong.
I think that a trader, when analyzing pairs, would be much better suited to look at regular correlation models or work on linear regression models where the Durbin Watson statistic shows little autocorrelation.
Sal Pellettieri, CFA, MA
Ron McEwan replies:
The Durbin Watson statistic is employed to determine if adjacent regression residuals are correlated. One of the assumptions of regression analysis is that the regression residuals are independent of each other. Sometimes, however, the data may unknowingly contain an "order effect," meaning that a previous measurement could influence the outcome of the successive observations.
The error terms of the ordinary least-squares equation estimate must be independently distributed of each other, and hence, the covariance between any pair of error or residual terms must be zero. Should this covariance be nonzero, then the residuals are said to be autocorrelated, and a relationship between present and past values can be found.
I agree with Mr. Pellettieri's comments that "autocorrelation is more an indication that the regression model being implemented is misspecified." In the article, I was not trying to identify the correct variables for a regression model. I wanted to exploit the idea that a "pair" of securities, where the covariance is nonzero, can reveal a predictive relationship. Using a simple correlation model may not always reveal the predictive relationship between the securities.
I do agree with Mr. Pellettieri that traders should apply alternative analytical routines to determine for themselves what works best for what they are trying to achieve.
CYBER CYCLES WITH INVERSE FISHER TRANSFORM
Editor,
I think John Ehlers thinks most traders want only the end results needed to trade (he may be right on this), plus his writing style is "compact," which leaves even traders desiring more underlying knowledge and understanding of how to use the end results. His results are great, but the details explaining where he gets his resultant equations sometimes are too sparse.
For example, take Ehlers' concept of the cyber cycle ("The Inverse Fisher Transform," May 2004 STOCKS & COMMODITIES). Ehlers says this code isolates the cyclic component via a filter. If we look at the code, we can see that if we assume alpha is zero, rather than 0.07, and use P for price in place of using a smoothed price, then we can better see the bare bones of the method he uses to develop an indicator for detecting cycles:
So in the first equation, we have a zero-sum weighting of 25%, -50%, 25% for the last three bars of price P; and in the second equation, a positive-sum weighting of 25%, 50%, -25% for the last three bars of the cycle.X = Cycle[0] = P - 2P[1] + P[2] Cycle[0] = X + 2Cycle[1] - Cycle[2]I would ask Mr. Ehlers to please explain the filtering method/technique used and how he arrived at this weighting scheme. Looking in his book Rocket Science For Traders, I suspect that it's related to the Hilbert transform, and inphase and quadrature components, since that is a primary method for discovering cyclicality from a time series, but I cannot find the proper parallel in the book that works with the article's text and code.
I'm guessing that the (1-alpha) and (1-alpha / 2) coefficients and their squares are somehow related to exponential smoothing, but again, please explain the basis for the code, as it is not at all clear.
Also, how did he quantitatively or theoretically pick 0.07 for alpha? Is it related to a minimum cyclic period that works with this indicator, or...?
William Smith
Cleveland, OHJohn Ehlers replies:
Thanks for your careful reading of my article. The purpose of the article was to expose readers to the inverse Fisher transform and to show some examples of how it might be applied to trading.
The inverse Fisher transform can be used with almost any oscillator that has been properly scaled. In my May 2004 article, I chose to use the RSI and my cybernetic cycle oscillator as examples of application. There is not room here to fully develop the cybernetic cycle concept, but I have done this derivation in chapter 2 of my latest book, Cybernetic Analysis For Stocks And Futures.
I hope you enjoy using this oscillator and other indicators and systems I derive in this book.
EXCEL FILE FOR VFI?
Editor,
Thanks for Markos Katsanos' article in the June 2004 issue of S&C ("Using Money Flow To Stay With The Trend"). He mentions at the end of the article that part 2 will be published in an upcoming issue. Could I suggest programming his VFI into an Excel spreadsheet? It would help in understanding exactly what he is doing.
Bob Dix
Boulder, COThe second article by Markos Katsanos appears in the July 2004 S&C. Katsanos again provided MetaStock code with his article, but also kindly provided us with an Excel spreadsheet file, which we have made available for download from our Subscribers' Area at https://technical.traders.com/sub/sublogin.asp. (Login requires your last name and subscriber account number, found on your magazine mailing label.)
We understand the benefit of presenting a technique in an Excel spreadsheet format because of the clear logic and accessibility of Excel. Unfortunately, it's not always possible to provide it, since that would be double the work for the author if he or she is already working in another program. It is our hope that no matter what kind of code is presented, it will at least help demonstrate the implementation of the technique.
In addition, Katsanos' article was again the topic for the Traders' Tips column in the July issue, so you will find many different program implementations of the technique. We hope that at least one of them will serve readers' needs. --Editor
MORE ON EXCEL CODE
Editor,
I am a longtime subscriber to your magazine and have found it to be an invaluable resource in my quest to become a better technical analyst. One section that I find particularly helpful is the Traders' Tips section, which helps readers program some of the indicators presented in the articles. However, I think one thing is missing: code for Microsoft Excel.
I feel that there are readers like me who would love to see Excel code presented in the Traders' Tips section. I've noticed that once in a while, a contributor of one of the articles will include Excel code in the article, but it is far and few between.
I use Excel quite a lot to do technical analysis and I find it an invaluable learning tool. When developing indicators one step at a time, in each column of a spreadsheet, the format of Excel really helps you gain a greater knowledge of the indicator you are programming and the mathematics behind it. With greater understanding comes greater creativity in developing indicators and trading systems.
Perhaps one or more of your staff writers who are knowledgeable in Excel could present the Excel code for the benefit of your readers. I believe this would be of great educational benefit to traders, both novice and advanced.
M. Carfi, via email
We appreciate your writing. Unfortunately, we can't codify all our authors' techniques in Excel, since the authors themselves would be the best ones to do that. Since they usually are already using another program to perform their research, we can't require them to use both. But we do encourage all our authors to include Excel spreadsheets implementing their techniques whenever possible. --Editor
VOLUME FLOW INDICATOR
Editor,
I enjoyed Markos Katsanos' June 2004 article, "Using Money Flow To Stay With The Trend."
I have a question for him. Instead of coloring the volume bars, has he looked at coloring the price bars? If so, what would the MetaStock formula be for this?
Jack Singer
Vancouver, BC, CanadaMarkos Katsanos replies:
I am glad you liked the article. Don't miss the second part in the July 2004 issue!
I included the formula for color-coded volume bars only because it will also curtail huge volume spikes, which tend to distort the scale of the volume histogram and make the rest of the volume bars barely visible.
To color-code price bars as well, open the expert advisor, click New, name the expert color-coded price bars, click Highlights, click New, name the new highlight "Up," choose green from the color drop-down list and paste the following formula:
INTER:=Log(Typical())-Log(Ref(Typical(),-1)); VINTER:=Stdev(INTER,30); CUTOFF:=.2*VINTER*C; MF:=Typical()-Ref(Typical(),-1); MF>CUTOFF
Similarly, click New again, name the new highlight "Down," choose red for color, and paste the following formula:
Good luck.INTER:=Log(Typical())-Log(Ref(Typical(),-1)); VINTER:=Stdev(INTER,30); CUTOFF:=.2*VINTER*C; MF:=Typical()-Ref(Typical(),-1); MF<-CUTOFF
VOLUME FLOW INDICATOR AND METASTOCK
Editor,
I copied your formula for VFI into MetaStock, but I think 130 days is too long; CKECK altria (MO).
What do you think of using a very short time (say, 13 days)? Is it representative?
Thanks for your answer and congratulations on your work.
Michel Laurent
BelgiumMarkos Katsanos replies:
I designed VFI as an alternative to on-balance volume (OBV), which is a long-term money flow indicator more for long-term investors.
If you are a short-term trader, the finite volume elements (FVE) indicator may be a more appropriate money flow indicator to use, which I discuss in the April 2003 and September 2003 issues of S&C.
The formula for FVE is:
PERIOD:= Input("PERIOD FOR FVE",5,80,22); COEF:=Input("COEF FOR CUTOFF",0,2,.1); INTRA:=Log(H)-Log(L); VINTRA:=Stdev(INTRA,PERIOD); INTER:=Log(Typical())-Log(Ref(Typical(),-1)); VINTER:=Stdev(INTER,PERIOD); CUTOFF:=COEF*(VINTER+VINTRA)*C; MF:=C-(H+L)/2+Typical()-Ref(Typical(),-1); FVE:=Sum(If(MF>CUTOFF, +V, If(MF <-CUTOFF, -V,0)),PERIOD)/Mov(V,PERIOD,S)/PERIOD*100; FVEFVE is an indicator similar to VFI, except that it takes into account intraday action as well. You can use a period from 12 to 28 days, according to your trading style.
VFI VINTER PERIOD
Editor,
Thank you for Markos Katsanos' June 2004 article on the volume flow indicator (VFI). It was a refreshing way to use volume as a trading tool.
I have a question: In the article, you mention that standard deviation of the change of the log of typical price was measured over a 30-day period. Did you mean a 30-day calendar period or a 30-day trading period? Most other measures of volatility (in the options world) use a 30-day calendar period, or a 21- or 22-day trading period. In the latter case, your MetaStock formula should be VINTER:= Stdev(INTER,22);
Dave Nowak, via email
Markos Katsanos replies:
The appropriate length of time over which to estimate the volatility is controversial, to say the least. Of course, standard deviation is not constant over time, and over longer periods (over 50 days), there can be substantial changes in the mean and standard deviation, as actual values of a financial price series are extremely unlikely to be normally distributed.
To calculate the historical volatility in my VFI formula, I chose a segment of 30 trading days, which is less than 50 but long enough to capture the short-term volatility.
This default value is not crucial, as it will only contribute to the threshold cutoff, and any value from 20 to 40 bars will make little, if any, difference to the final formula result.
ERRATA: INVESTOR/RT
Editor,
In a letter to the editor in the April 2004 issue of S&C, a reader asks about charting software that runs on Mac OSX. Your reply was somewhat surprising.
You mentioned that the most recently reviewed Macintosh software was Behold, in the January 2002 issue. In fact, the most recently reviewed Macintosh software was Investor/RT in your August 2003 issue. Furthermore, Investor/RT is available as a native Macintosh OSX application, and has been for several years, as pointed out by your review last August.
Although Investor/RT has the distinction of being the only software of its kind that runs equally well under MS Windows, Macintosh Classic, and Macintosh OSX, the product was designed from inception to operate natively on the Mac platform. Investor/RT does compete in the MS Windows software marketplace as well, and some of our clients are, in fact, former MetaStock software users.
Your readers interested in advanced charting and trading system development software for Macintosh OSX can learn more about Investor/RT at www.linnsoft.com. A link to your August 2003 review of Investor/RT is available there as well.
William E. Linn Jr.
President
Linn Software, Inc.We did indeed review Investor/RT 6.15 in our August 2003 issue. We apologize for this oversight. Thank you for correcting this point.-Editor
ERRATA: OVERNIGHT TRADING
Two figures in Anthony Trongone's July 2004 article, "Overnight Trading," were printed incorrectly. On page 20, Figure 3B shows an incorrect chart, and on page 5, some of the up and down arrows in the table in Figure 5 didn't print. The correct figures for 3A, 3B, and 5 are shown on this page. We regret this error. --Editor
ERRATA: MATT BLACKMAN CONTACT INFORMATION
At the end of Matt Blackman's June 2004 article, "Island Trader," we inadvertently printed outdated contact information and biographical information for this author. Here is his updated information:
Matt Blackman is a technical trader, author, reviewer, keynote speaker, and regular contributor to a number of trading publications and investment/trading websites in North America and Europe. He is an affiliate of the Market Technicians Association and a member of the Canadian Society of Technical Analysts, and he is currently enrolled in the Chartered Market Technicians (CMT) program. He can be reached at matt@tradesystemguru.com or trader@goldhaven.com.
ERRATA: APRIL 2004 TRADERS' TIPS
Editor,
I have a correction to our code given in the MetaStock Traders' Tips column from the April 2004 S&C. This corrects a problem in the code for the B-indicator.
In trying to match the chart presented in the article on which the tip was based ("Trend-Quality Indicator" by David Sepiashvili), I had the formulas in the original code using a seven- and 15-period moving average, as was done on the Q-indicator chart in the article. However, the text in the article says they are actually using 10 and 40 periods for that chart, and that turned out to be the problem.
Here is the corrected MetaStock formula for the B-indicator:
m:=Input("% Scalar trend period",1,25,4); n:=Input("% Scalar noise period",1,500,250); cf:=Input("% Scalar correction factor",1,250,2); p1:=Input("First moving average periods",1,200,7); p2:=Input("Second moving average periods",1,200,15); rev:=Mov(C,p1,E)-Mov(C,p2,E); pds:=If(rev>0,1,-1); dc:=ROC(C,1,$); cpc:=If(pds<>Ref(pds,-1),0,(dc*pds)+PREV); trend:=If(pds<>Ref(pds,-1),0,(cpc*(1/m))+(PREV*(1-(1/m)))); dt:=cpc-trend; noise:=cf*Sqrt(Mov(dt*dt,n,S)); temp:=If(Abs(trend)+Abs(noise)=0,1,Abs(trend) +Abs(noise)); (Abs(trend)/temp)*100;In addition, here is the corrected code for the Q-indicator. This was posted earlier, but to be safe I'd like to give it again:
m:=Input("% Scalar trend period",1,25,4); n:=Input("% Scalar noise period",1,500,250); cf:=Input("% Scalar correction factor",1,250,2); p1:=Input("First moving average periods",1,200,7); p2:=Input("Second moving average periods",1,200,15); rev:=Mov(C,p1,E)-Mov(C,p2,E); pds:=If(rev>0,1,-1); dc:=ROC(C,1,$); cpc:=If(pds<>Ref(pds,-1),0,(dc*pds)+PREV); trend:=If(pds<>Ref(pds,-1),0,(cpc*(1/m))+(PREV*(1-(1/m)))); dt:=cpc-trend; noise:=cf*Sqrt(Mov(dt*dt,n,S)); trend/noise
William Golson
Equis Technical Support
www.MetaStock.comThank you for providing this updated code. We have also posted this code at our website in the April Traders' Tips area at https://www.traders.com/Documentation/FEEDbk_docs/Archive/042004/TradersTips/TradersTips.html. --Editor
KELTNER CHANNELS
Editor,
I am writing a book about trading with Keltner channels. By way of research for an early chapter, I'm trying to find anyone who knew Chester Keltner so I could ask some questions about him.
Do you know of anyone, particularly grain traders, who may have known Keltner first-hand? If so, can you refer me to him/her?
Vince Heiker, via email
Flower Mound, TXI cannot think of anyone to refer you to who knew Chester Keltner. Perhaps someone reading this will have a suggestion.
However, based on the backtesting we published in the December 1999 S&C, STARC bands and Bollinger Bands deserve a comparative look. The 10-day moving average rule (Keltner channels) described in How To Make Money In Commodities by Chester Keltner [1960] seems less promising.
Good luck on your book. --Publisher
Back to August 2004 Contents