Gold probably generates more diverse opinions than any other financial instrument or commodity. The deflationists say gold is going down to $250 while the inflationists say gold is going to $5,000 and beyond. The price action paints a much simpler picture from the cyclical/spectral standpoint. The first chart below is a monthly chart showing what appears to be a dominant eight year price wave. The filter does a good job of identifying both the highs and the lows of that wave. Trying to phase a wave of that length to a particular day at nowtime is an exercise in futility because of the relatively large margin of error due to filter lag. Nevertheless, if one projects the average period from the most recent low in 2008 into the future, the next low of that wave is due sometime in 2016 barring a large duration fluctuation. Additionally, at each of the prior 8 year lows the highpass data exhibited a monthly close below -1.5 average deviation (red horizontal line), indicating the maximum distance the 8 year wave is going to push it (somewhat analogous to an FLD projection but in the spectral domain).
Gold also exhibits some very clear intermediate price waves. The chart below is a weekly chart illustrating two of those waves. The longer wave at the bottom of the chart had a tendency over the last several years to phase to cyclical highs, (yellow arrows), a characteristic that has been discussed recently by other analysts. The shorter wave has been very dominant and consistent over the past seven years. If that dominancy and consistency persists over the next year or so, identifying the window of the next major low should be relatively easy.
There is a caveat to the above analysis. The waves shown, in terms of average period and their harmonic relationship to one another, do not fit into the standard Hurst simple harmonic, time domain model. I do not subscribe to the approach that one size (nominal model) fits all. I analyze each instrument individually to determine the character of the dominant price waves. A standard Hurst analysis of gold using his nominal model should not be that much different.
7 thoughts on “Goldbugs”
Hi William,
The precious metal sector is an interesting sector. Lots of volatility in both bullish and bearish markets. I have traded this market for some time. I posted in early 2013 on the original Sentient Trading site of Gold’s secular high (=1947 high).
http://hurstcycles.com/7929-2/#sthash.8yi16y8V.dpbs
I too follow an 8 year cycle. For gold I have 15.8 yr, 7.9 yr, and 47 months. The issue I always have is the next harmonic, ie. is it approx. 16 month (x3), or 12 months (x4)? Your bandpass filter work suggests the latter. That does not surprise me as a 6 month cycle is the most tradeable cycle and the one that I have followed for a while now. As you suggest, gold is due an important nest of lows and a sustained rally is due from that low. In gold’s bear market from 2011, the gold miners have not mnaged to rally past 12 weeks before setting new lows or completely retracing the swing up. I am looking for some very tradeable rallies in the future. BTW silver for some reason seems to run a slightly different cycle (I have it a slightly shorter cycle periods).
I follow many aspects in the PM space to try to keep on the right side of the market. The Gold COT report is one of them and judging from the position (extreme short at this point) of the commercials, this is not a time to be bullish, but cautious.
cheers,
john
Hi John.
Many commodities (and forex) have dominant waves that exhibit a complex harmonic relationship to one another. From a spectral (i.e. bandpass) viewpoint it is a non-issue but it can cause difficulties if one’s phasing analysis is based on a simple harmonic relationship in the time domain. For example the two waves on my intermediate chart currently have a 14 to 4 relationship! Some might contend that the filter output is incorrect because they don’t exhibit the expected 4 to 1 relationship, but the math and the eyeball test say otherwise. Even the longer intermediate wave and the 8 year wave exhibit an approximate 5 to 1 harmonic relationship! Don’t confine yourself solely to the simple harmonic relationship box.
I noticed in your COT chart that the large speculators and the commercial speculators have an almost perfect inverse relationship to one another and they in turn tract very closely to the dominant wave shown on the daily chart below. You will notice that the wave below has a 3 to 1 harmonic relationship with the shorter wave on my intermediate chart which means that the waves will synchronize at highs and lows. This reduces the transaction risk to a very low level. I reiterate that one should let the price action dictate what is dominant and what the harmonic relationship is as opposed to imposing an artificial structure that may or may not be accurate.
William
William,
I really appreciate your posting your bandpass filter work with those different views. I understand what you are saying wrt harmonics. But I have always tried to follow a 26/28 week cycle (+/-) and a 13/14 week cycle. This has been mostly from staring at charts for a long time and following this sector for many years. It has always been a struggle trying to move these cycles up to the larger 4 and 8 year cycle periods which have always been fairly obvious in their influence. Just eyeballing everything, it seem to me that an important bottom could well arrive by October this year. But as you say price action rules all. I will be looking closely for that long term swing entry point.
cheers,
john
Hi William and John
Firstly, I’d like to thank you both for the knowledge that you have shared in this community. William, I would especially like to thank you for highlighting the spectral domain and its importance in Hurst’s work.
In our application of Hurst analysis we default to using what has become known as the Hurst Nominal model which is a collection of harmonically related periods of time that Hurst found present in his work on equity prices. But as William has pointed out, Hurst derived his time related Nominal Model from his spectral model, which he derived, not in the time domain but in the frequency domain. Hurst then selected groups of angular frequencies that were related and labeled them with the periods of time they best corresponded. These periods of time represent ranges and are not exact finite numbers. Additionally, the concept of synchronized troughs and perfectly harmonically related cycles are tendencies, not inviolate laws….the real world is not so neat. To make matters even more challenging, these cycles come and go and expand and contract. Fortunately, he showed that this happens slowly.
In light of the above, I believe that it makes sense to determine what the dominant market vibration has been and begin our analysis from this information ie: extract information from the data and only then use this information in a phasing analysis. This is the opposite approach to taking the Hurst Nominal Model and applying it to the data
William has done exactly what Hurst did: extract cyclic information through the use of filtration. Hurst also used Fourier Analysis to deconstruct market data. The Goertzel algorithm allows for evaluation of individual terms of the analysis efficiently and accurately. Dennis Meyers (www.meyersanalytics.com) and Lars Von Thienen (whentotrade.com) have done important work to adapt the algorithm to market data.
In the attached monthly chart, the 93 mo vibration is dominant with the 48 month vibration also present in the data. This correlates with William’s bandpass information. Note also that the peaks of both of these cycles are correlated with the peaks in the data with the trough correlation being less convincing.
The attached Sentient trader chart shows the phasing analysis using a custom nominal model incorporating the 93 month vibration and synchronized peaks.
What is interesting is that with only only input of 93 months, the actual nominal model Sentient Trader has calculated has phased the 4.5 year cycle at 46.8 months agreeing +/- 1 month with the Goertzel output of 48 months.
Repeating the process with the weekly data, we note a 27 week cycle dating from the 1970’s morphing into a 62 week cycle at the 1999 low. 62 weeks(15.5 months) is very close to what Sentient has found for the period of the 18 mo cycle..
The final chart is a close up of the longer term phasing the inclusion of the three longest cycles independently extracted from almost 40 years of gold time and price data
first chart
second
third