
Noted precious metals investment expert Michael Maloney is no stranger to bold statements and even bolder predictions. Those familiar with Maloney's work will undoubtedly recall his multiple arguments about a deflationary crash in the Dow Jones and an inflationary explosion in the price of bullion. While the case for $10,000 gold and beyond will ultimately be decided by its culmination, or lack thereof, there is another point Maloney made that deserves special consideration.
Back in 2010, Mr. Maloney was invited to speak at the 8th International Banking Forum in Sochi, Russia. There, he gave his usual expose of rampant monetary expansionism by global central banks and an impending sovereign debt and currency crisis. In order to prove some of his heavy accusations, he cited the discipline of technical analysis, specifically mentioning the occurrence of "head-and-shoulders" patterns (part of a sub-sector of technical analysis called Elliot Wave Theory) as a harbinger of bearish events in global equity markets.
Citing technical analysis is nothing new but assigning a nominal probability range for its accuracy is. Technical analysis has always suffered from a "meritocracy" problem by which its calculated implications can only be asserted on a subjective basis. That is, the discipline's usefulness is palpable but sadly unquantifiable. This prevents the technical approach from achieving scientific credibility and will likely never attain such a status.
So it was a bit of a surprise when Maloney in his speech blurted out that technical analysis has a success rate of 55- to 60- percent! There is no doubt that he pulled this number out of his ass (or that he cited someone who pulled this number out of his ass). No entity to my knowledge has ever performed an empirical study on the merits of technical analysis. Further, such a study is made impossible since there is no concrete consensus on what exactly constitutes technical analysis.
However, Maloney should not be blamed for announcing this assertion. While the technical approach lacks scientific standing, that by itself is not reason to dismiss it. Even the widely accepted discipline of accounting lacks such status, and is thus considered an "art." Also, that technical analysis "works" is starting to gain more mainstream acceptance. Media outlets such as CNBC regularly examine select stocks from a fundamental and technical perspective.
But no one has bothered to empirically check the discipline's validity...until now!
Writing a simple program, I calculated the nominal parameters of what constitutes a head-and-shoulders pattern, the key subject of Maloney's discussion on technical analysis. A big part of the reason why no one else had undertaken the task of empirically testing the technical approach is the logistics of the project : no one is going to manually rifle through pages and pages of price charts looking for specific patterns and formations. For such an undertaking, an algorithm is necessary.
Upon completion of my program, I ran it through the entire history of the S&P 500 index, noting when a head-and-shoulders pattern occurred and the percentage by which such a pattern led to a decline in market value. I also ran the exact same experiment for an "inverse" head-and-shoulders pattern, its twin model that implies a bullish future for market valuation.
To my surprise, the head-and-shoulders pattern was remarkably accurate, producing a bearish outcome 67.82% of the time. The inverse variety was even more accurate, producing a bullish probability of 81.3%. This confirmed to me that the U.S. equities market has an upward bias ; indeed, a separate study that I conducted shows that on any given day, the market will produce a positive result more than 52% of the time.
Since we are on the topic of Michael Maloney, I decided to conduct the same experiment with the popular GLD, an exchange-traded fund that tracks the gold market. Here, the head-and-shoulders pattern produced a bearish result 66.82% of the time, a minute 1.49% variance against the S&P 500. But a substantial difference occurred when comparing the inverse pattern, with the GLD only achieved a bullish probability of 79.17%. While this may seem small, the variance of the inverse pattern in the GLD is almost twice that of its equivalent in the S&P index.
On surface level, this tells us that an inverse head-and-shoulders pattern is slightly more likely to "fail" than an equivalent pattern appearing in the equities market. Such an observation may embolden the gold and silver bugs who adamantly claim that the bullion market is rigged. From a non-partial perspective, it is interesting and worthy of further research. The variance between the inverse results of the two markets is in my opinion statistically significant ; however, on a pure nominal basis, it will be difficult to condemn an entire market as being manipulated because of a 2% discrepancy in a discipline that can often produce more questions than answers.
Even my own analysis is subject to criticism. Even though the methodologies that I have applied were consistent throughout the analysis, some of the "hits" that my algorithm found would not be considered a true head-and-shoulders pattern. This meets an input dilemma (as the programmer, I had to define to the computer what a head-and-shoulders pattern is) and an empirical dilemma (nobody knows what a head-and-shoulders pattern really is). Using statistical language, my degree of error is unknown. Ironically, an attempt to quantify the degree of error will lead to an existential discussion and an inordinate amount of academic bickering, a journey that I just plainly refuse to undertake.
Still, the consistency of methodology over several years of data does confirm semi-empirically what technicians have always believed : technical analysis is accurate. And when performed by the right people, it may be more accurate than you may be giving it credit for!