Technical analysis in major brokerages and financial media

Weather prediction, medical diagnosis and technical analysis

Suppose, in the weather forecast part of a local newscast, the person handling the weather displays a chart of recent temperatures in the local area, points out “trends” and “waves,” then mentions a “breakout pattern” from a recent temperature range. Most of us would not have much confidence in such a dubious and unorthodox forecast, and, if followed (e.g., for a major storm), could have serious consequences.

Or suppose, when one’s electrocardiogram is taken at a clinic, that the attending physician makes some measurements by hand between some events on the graph and notes a “Fibonacci ratio” between them [the “Fibonacci ratio,” also known as the “golden ratio,” usually denotes either the value (sqrt (5) + 1)/2 = 1.6180339… or its reciprocal, namely (sqrt(5) – 1)/2 = 0.6180339…]. Or suppose that after reviewing one’s historical record of pulse and blood pressure measurements that the physician noted a “triangle pattern” in the data. If such were to happen, a patient would be entirely justified to discontinue seeing this physician.

The reason that scenarios such as this do not happen in most scientific disciplines and professions is because in recent years most fields have significantly upgraded their prevailing standards for data analysis. Typical of these upgraded standards is this statement by the American Statistical Association on the usage of p-values and significance statistics. This statement takes aim at “p-hacking,” namely the ethically questionable practice of testing numerous hypotheses on a single set of data until one finds a hypothesis that is confirmed to, say, the level p = 0.05. See this Mathematical Investor blog: P-hacking and backtest overfitting and this Math Scholar blog: P-hacking and scientific reproducibility for additional details.

Technical analysis in major brokerages and financial media

So what is one to think about the numerous major brokerages and financial media outlets who continue to promote technical analysis, including “trends,” “waves,” “breakout patterns,” “triangle patterns” and “Fibonacci ratios”? In a previous Mathematical Investor blog The most important plot in finance, we mentioned some of these organizations. Here are some more details, with updates:

  1. Charles Schwab represents technical analysis as an indispensable tool for active traders here and here. At the first URL, for instance, one reads

    This chart shows five short trends higher and five short trends lower. Notice that as price declines, volume rises and as price rises, volume declines. This is exactly the opposite of what a trader would want to see in a rising or “bull” market and typical of a falling or “bear” market. … In this type of situation, consider using tighter-than-normal stop-loss orders to protect capital and/or profits against the possible resumption of the longer-term downtrend…

  2. Merrill Lynch offers a Market Analysis Technical Handbook. Among the chapter headings are “Price Momentum Indicators,” “Support and Resistance,” and “The Fibonacci Concept.” Yes, one chapter is devoted to the “Fibonacci concept.”
  3. Some Bank of America / Merrill Lynch analysts utilize technical analysis, for instance here. Note that this analyst cites a “double breakout” pattern, which led him to believe that the bond market would soon rally.
  4. Fidelity considers technical analysis an appropriate technique for both individual and professional traders, as shown here and here. At the latter URL one reads that their Technical Indicator Guide “can help you identify possible entry and exit points for trades and may help you achieve your investing goals.”
  5. E-Trade offers an introduction to technical analysis here. In this introduction, one learns that technical analysis is built on several assumptions, two of which are “stock prices tend to move in trends” and “history repeats in the stock market.”
  6. Barrons explains here why they believe technical analysis matters:

    [I]f a stock is rising and then starts to move sideways as bulls and bears become uncertain as to what to do next, a coiling pattern appears on the chart as price swings in both directions diminish. … Chart watchers wait for prices to move above the upper border of the triangle and then buy the stock because the odds favor further gains.

  7. The Wall Street Journal lists numerous articles on technical analysis here. In this article one reads “Stocks can’t rise forever, but the bad news for bears is that few clear ‘sell’ signals have materialized, technicians say.”
  8. Bloomberg offers a blog dedicated to technical analysis here. This blog quotes an analyst as saying that the best definition for technical analysis is “using technology to improve investment results.”
  9. MarketWatch.com offers a blog with a promotion of technical analysis tools here:

    Using candlestick charts and proprietary tools, [the toolkit] establishes near-term market bias and identifies patterns, trends, support and resistance levels, moving averages, attractive entry and exit points, buying opportunities and more.

  10. The Market Realist recommends technical analysis and Elliott wave theory as investment tools here. It explains:

    In the beginning of the corrective phase, it’s difficult to predict that the market trend has stopped. The stock price starts dropping due to correction. This phase is called wave A. In wave B, the stock price increases on the anticipation that the long uptrend is still there. Wave C is formed when stock market news is negative and a downtrend is confirmed.

Why technical analysis doesn’t work

In some previous Mathematical Investor blogs, we have cited several reasons why technical analysis does not work and cannot be expected to work. Here is a summary:

  1. Disagreements. One difficulty is that there is no clear, publicly acknowledged consensus within the technical analysis community as to what constitutes an actionable pattern, particularly when chart readings and interpretation rules are themselves vague and ambiguous.

  2. Research studies by the present authors. One of the present authors and colleagues recently completed the study Evaluation and ranking of market forecasters, which analyzed the records of 68 market forecasters, based on data earlier collected by CXO Advisory, and employing a novel weighting scheme that took into account how specific the forecasts were. Among these 68 forecasters were 27 who acknowledge using technical analysts as a significant part of their analysis. So how well did these 27 technical analysts do? Their average precision score was 44.1% — in other words, less than even chance. In fact, this average score was slightly less than the average of all 68 forecasters in our study. In short, there is no evidence whatsoever in this data that technical analysis is effective in predicting markets. If anything, our results must be on the optimistic side, because of the well-known survivorship bias phenomenon — very likely numerous unsuccessful technical analysis practitioners have dropped out of the business, and thus are absent from our tables.
  3. Research studies by other analysts. We are hardly alone in concluding that technical analysis does not work. Market analyst Laszlo Birinyi, for instance, interviewed in the book The Heretics of Finance, declared, rather bluntly, “The truth is technical analysis doesn’t work in the market.”
  4. Multiple testing errors and cherry picking biases. Technical analysts can cite some successes, but statistically speaking, how “real” are these? As we showed in the earlier Mathematical Investor blog The most important plot in finance, many of these claims fall prey to multiple testing errors and cherry picking biases. And as we mentioned in the introduction above, the American Statistical Association recently released a statement condemning “p-hacking,” which is exactly the same practice — trying numerous hypotheses, or searching thousands or millions of possibilities on a computer, and only highlighting or publishing the one or a handful of cases that look the best. This is the essence of backtest overfitting in finance, which sadly afflicts many arenas of the field — see P-hacking and backtest overfitting.
  5. Fundamental considerations of the operation of markets. Numerous quant funds and other organizations at the forefront of modern quantitative finance employ highly sophisticated mathematical algorithms (much, much more sophisticated and extensive than anything ever used in the technical analysis world), with huge dynamic datasets, implemented on state-of-the-art large-scale computer equipment, and trading at millisecond and even microsecond levels. What’s more, increasingly these are the only organizations that consistently make money — see, for instance, Majority of highest-earning hedge fund managers and traders are at quant firms. Furthermore, these programs are engaged in a very real “arms race,” because any strategy that works is quickly mimicked by other programs from other organizations, so that any “edge” evaporates rather rapidly. Indeed, this “war” partly explains why the resulting market price stream is almost entirely a random walk. So those who promote technical analysis would have us believe that the many highly trained mathematicians and their highly sophisticated computer programs have all somehow missed a few simple, elementary schemes that anyone armed with a laptop, a plotting program and a handful of simple tools can routinely take advantage of to produce reliable above-market-average profits. Obviously there cannot be any such trivial schemes.

Why this matters

The technical analysis community certainly has command of the financial news world — it is hard to read any online financial news source without seeing at least some articles of this genre. More importantly, many individual investors, pension funds, mutual funds and other organizations routinely use technical analysis methods in their market analysis and decision making. As the quotes in the brokerage/media section above declare, technical analysis is “using technology to improve investment results,” and “can help you identify possible entry and exit points for trades and may help you achieve your investing goals.”

And therein lies the problem. Hundreds of thousands or more of investors, both individual and professional, have been firmly convinced that technical analysis is how the “smart” investor wins in the market.

As one of us wrote in this Forbes interview, “[Technical analysis strategies] encourage traders and investors to put their money to work, while offering guidance with no objective information value.” If similarly ineffective data analysis techniques were employed in the medical or pharmaceutical world, lawsuits would be lodged.

What’s more, as the quote about “entry and exit points” makes clear, market timing is central to using technical analysis methods. But the overwhelming consensus of professional analysts is that market timing is a VERY BAD STRATEGY, especially for individual investors with retirement accounts. In addition to the extra fees paid, in all too many cases investors sell out significant portions of their portfolio in a panic just before or after the market prices hit bottom, only then to miss a substantial rebound. For example, a financial advisor known to one of the present authors reports that at least one client who exited the market at the wrong time in the wake of the 2008-2009 downturn never recovered his/her previous level of capital. Other clients have been convinced to go “all in” during market highs, and yet have been reluctant to buy during market lows.

Such failures of market timing are multiplied by the thousands and tens of thousands in the investment world, and account for a significant fraction of the chronic poor performance of individual investors in particular — see these Mathematical Investor blogs: Poor investor performance: What can be done? and The folly of panic selling.

Hope on the horizon

There is some hope on the horizon. The success of state-of-the-art quantitative finance and machine learning methods is now more widely recognized in the field, and there are some indications that even the technical analysis community is starting to explore some of these more advanced schemes. Analysts with real-world quant and machine learning skills are in high demand — see What is the best training for finance PhDs. Along this line, a colleague of one of the present authors notes:

At one time, when I started doing recruitment at [a hedge fund], it was common — and acceptable — for candidates to tout their technical analysis skills. Now, at the various firms where I have worked, that would pretty well disqualify them on the spot. When I recently spoke to a national group of active investment managers, I saw a similar shift in a constructive direction.

But it is also important, just for the credibility of the mathematical finance field, that more of us are willing to declare “the emperor has no clothes” when we see material advocating techniques that are clearly ineffective and/or out of date. As one of us and colleagues explained in a 2014 paper, Our silence is consent, making us accomplices in these abuses.

Comments are closed.