If you torture the data long enough, it will confess.


What’s so hard about predicting stock prices? Anyone with open eyes can see patterns in stock prices—patterns that are a self-evident foretelling of whether prices are headed up or down.

Thirty years ago, a former student named Jeff called me with exciting news—my lectures on the futility of trying to discern profitable patterns in stock prices were hogwash. Jeff had taken a job with IBM and, in his spare time (ha, ha!), was studying stock prices and had found some clear patterns. He was fine-tuning his system and would soon be rich. He told me that he was going to rent a helicopter and land it on the lawn outside my classroom so that he could march into my investments class triumphantly and tell students the truth.

Every year, I tell the students this story. Then I walk over to a classroom window and look outside to see if Jeff’s helicopter is parked outside. I’m still waiting.


Technical analysts try to gauge investor sentiment by studying stock prices, trading volume, and various measures of investor sentiment. Technicians do not look at dividends or profits, either for individual companies or for the market as a whole. If they are studying an individual company, they don’t even need to know the company’s name. It might bias their reading of the charts.

A technical analyst can be compared to a person who watches a computer program draw lines on a monitor and tries to discover a pattern that will predict the next line to be drawn. The lines themselves are all that matters and it would be distracting to think about whether the computer program was written in Java or C++. In the same way, the mood of the stock market can be gauged by watching stock prices; additional news about the economy or specific companies would be distracting. John Magee, who coauthored the so-called bible of technical analysis, boarded up the windows of his office so that his readings of the hopes and fears of the market would not be influenced by the sight of birds singing or snow falling.

A technician’s most important tool is a chart of stock prices. The most popular are vertical-line charts, traditionally using daily price data. Each vertical line spans the high and low prices, with horizontal slashes showing the opening and closing prices. A technician adds lines, like the channel in Figure 4-1, to show a trend or other exploitable pattern (to reduce the clutter, I omitted the vertical lines and just show the closing prices).

FIGURE 4-1. A graph of closing prices reveals a channel



Humans have a natural affinity for professed experts who replace confusion and ambiguity with clarity and decisiveness. Periodically, a technical analyst is elevated to the status of financial guru when astoundingly accurate predictions are reported in the media and devoted followers seek the advice of these celebrities.

For example, Joseph Granville was a flamboyant guru, sometimes enlivening his public speeches with vaudeville skits using a chimpanzee or ventriloquist’s dummy and at other times preaching in a prophet’s robes:

The market is a jealous God. It rewards winners and chastises losers. The Holy Bible is a record of winners and losers. The market follows every precept in that Book—if the market does not follow man’s ways, what does it follow? God’s ways.

His forecasts were so inaccurate in the early 1970s (due, he says, to an addiction to golf) that he abandoned the stock market completely. But then “golfers anonymous” turned his life around and he wrote a book, How to Win at Bingo, that sold 500,000 copies. Buoyed by this success and four years of good stock market predictions in the late 1970s, he boasted of having “cracked the secret of markets,” promised that he would never make a serious mistake again, and nominated himself for the Nobel Prize in Economics. In his spare time, he predicted that Los Angeles would be destroyed by an 8.3 earthquake in May 1981.

Granville issued a buy recommendation in April 1980 and the Dow jumped 30 points the next day, perhaps propelled by avid subscribers following his advice. In September, he continued to predict a runaway stock market. “Short sellers are about to get the heat, and if you think hell is hot, watch.” On January 6, 1981, he abruptly turned bearish; his post-midnight calls to his biggest subscribers telling them to sell all of their stocks (even while his normal market letter advised smaller subscribers to buy) provoked the “Granville crash,” a 24-point drop in the Dow on January 7. In his March 7, 1981, newsletter, he declared that the “March Massacre” had begun and that he couldn’t possibly be wrong in his prediction that the Dow would fall at least 100 points and possibly 200 points by May 1. In fact, the Dow rose from 965 to 996. His recycled sell recommendation in September 1981 sent tremors through financial markets.

In late September 1981, he appeared on British television, advising investors to sell everything. He predicted that British interest rates would jump from 14 percent to above 17 percent and that the British industrial stock index would sink like a stone, from 480 to 150. British interest rates peaked the next month and headed downward to 12 percent in 1982 and 10 percent in 1983. The stock index bottomed at 474 in October 1981, too, and headed upward, to 580, a year later, then 700 the year after, and then to 900 in 1984. It reached 1000 in 1985 and 1300 in 1986.

Meanwhile, back in the United States, Granville remained bearish throughout 1982–1985, missing one of the greatest bull markets. In 1982 and 1983, he advised his subscribers to sell short; those who did lost 30 percent in 1982 and then another 25 percent in 1983 as the market rose sharply. For 1984 he cited “333 exact parallels with 1929” and predicted that a comparable crash would drive the Dow Jones Industrial Average below 700 by the spring of 1985. The Dow topped 1,300 that spring and Granville stubbornly continued predicting a Second Great Crash.

The Hulbert Financial Digest, which tracks the performance of investment newsletters, put Granville dead last over the twenty-five-year period 1980 to 2005, with an annual return of –20 percent while the S&P returned 14 percent a year.


To many, the value of technical analysis is self-evident. Any reasonably alert person can see well-defined patterns in stock prices. However, investors need a crystal ball, and stock charts provide a rearview mirror. In any set of data, even randomly generated data, it is possible to find a pattern if one looks long enough.

Ransacking data for patterns is called data mining and demonstrates little more than the researcher’s persistence. Remember this chapter’s opening quotation: “If you torture the data long enough, it will confess.”

I once sent ten different charts of stock prices (including Figure 4-1) to a technical analyst—let’s call him Ed—and asked his help in deciding whether any of these stocks looked like promising investments.

Ed was so excited by the patterns he found in four charts that he overlooked the odd coincidence that all of the price charts started at a price of $50 a share. This was not a coincidence.

These were not real stocks. I created fictitious data from student coin flips. In each case, the “price” started at $50 and then each day’s price change was determined by twenty-five coin flips, with the price going up 50 cents if the coin landed heads and going down 50 cents if the coin landed tails. For example, fourteen heads and eleven tails would be a $1.50 increase that day. After generating dozens of charts, I sent ten of them to Ed with the expectation that he would find seductive patterns. Sure enough, he did.

When this ruse was revealed, Ed was disappointed that these were not real stocks, with real opportunities for profitable buying and selling. However, the lesson he drew from this hoax was quite different from what I intended. Ed concluded that it is possible to use technical analysis to predict coin flips!


Millions of investors have spent billions of hours trying to discover a formula for beating the stock market. It is not surprising that some have stumbled on rules that explain the past remarkably well but are unsuccessful in predicting the future. Many such systems would be laughable, except for the fact that people believe in them.

Analysts have monitored sunspots, the water level of the Great Lakes, and sales of aspirin and yellow paint. Some believe that the market does especially well in years ending in five—1975, 1985, and so on—while others argue that years ending in eight are best. Burton Crane, a longtime New York Times financial columnist, reported that a man “ran a fairly successful investment advisory service based in his ‘readings’ of the comic strips in The New York Sun.” Money magazine once reported that a Minneapolis stockbroker selected stocks by spreading the Wall Street Journal on the floor and buying the stock touched by the first nail on the right paw of his golden retriever. The fact that he thought this would attract investors says something about him—and his customers.


On Super Bowl Sunday in January 1983, both the business and sports sections of the Los Angeles Times carried articles on the Super Bowl stock market predictor. The theory is that the stock market goes up if the National Football Conference (NFC) or a former National Football League (NFL) team now in the American Football Conference (AFC) wins the Super Bowl; the market goes down otherwise. A Green Bay Packer win is good for the stock market; a New York Jets win is bad for stocks.

This theory had been correct for fifteen of the first sixteen Super Bowls, and one stockbroker said that “market observers will be glued to their TV screens . . . it will be hard to ignore an S&P indicator with an accuracy quotient that’s greater than 94 percent.” Washington (an NFC team) won, the stock market went up, and the Super Bowl Indicator was back in the news the next year, stronger than ever. The Super Bowl system worked an impressive twenty-eight out of thirty-one times through 1997, but then failed eight of the next fourteen years.

The stock market has nothing to do with the outcome of a football game. The accuracy of the Super Bowl Indicator is nothing more than an amusing coincidence fueled by the fact that the stock market usually goes up and the NFC usually wins the Super Bowl. The correlation is made more impressive by the gimmick of counting the Pittsburgh Steelers, an AFC team, as an NFC team. The excuse is that Pittsburgh once was in the NFL; the real reason is that Pittsburgh won the Super Bowl several times when the stock market went up. Counting Pittsburgh as an NFC team twists the data to support this cockamamie theory.

The New York Times reversed the direction of the prediction. Instead of using the Super Bowl to predict the stock market, why not use the stock market to predict the Super Bowl? Why not? It’s no more ridiculous than the original Super Bowl Indicator. The Times reported that, if the Dow increases between the end of November and the time of the Super Bowl, the football team whose city comes second alphabetically usually wins. (Hint: Why do you suppose they chose the end of November for the starting date, as opposed to January 1, a month before the game, a year before the game, or another logical date?)

The performance of the Super Bowl Indicator has been mediocre since its discovery—which is unsurprising since there was nothing behind it but coincidence. What is genuinely surprising is that many people do not get the joke. The man who created the Super Bowl Indicator intended it to be a humorous way of demonstrating that correlation does not imply causation. He was flabbergasted when people started taking it seriously!


In 1996, two brothers, Tom and David Gardner, wrote a wildly popular book with the beguiling name, The Motley Fool Investment Guide: How the Fools Beat Wall Street’s Wise Men and How You Can Too. Hey, if fools can beat the market, so can we all.

The Gardners recommended the Foolish Four Strategy. They claimed that during the years 1973–1993, this strategy had an annual average return of 25 percent and concluded that it “should grant its fans the same 25 percent annualized returns going forward that it has served up in the past.”

Here’s their recipe for investment riches:

1.At the beginning of the year, calculate the dividend yield for each of the thirty stocks in the Dow Jones Industrial Average. For example, on December 31, 2013, Coca-Cola stock had a price of $41.31 per share and paid an annual dividend of $1.12 per share. Coke’s dividend yield was $1.12/$41.31 = 0.0271, or 2.71 percent.

2.Of the thirty Dow stocks, identify the ten stocks with the highest dividend yields.

3.Of these ten stocks, choose the five stocks with the lowest price per share.

4.Of these five stocks, cross out the stock with the lowest price.

5.Invest 40 percent of your wealth in the stock with the next lowest price.

6.Invest 20 percent of your wealth in each of the other three stocks.

No, I’m not making this up.

Any guesses why this strategy is so complicated, verging on baffling? Data mining perhaps?

Steps 1 and 2 are plausible. There is a long-established investment strategy called the Dogs of the Dow that favors buying the Dow stocks with the highest dividend yields, and this sensible strategy has been reasonably successful.

But beyond this kernel of a borrowed idea, the Foolish Four Strategy is pure data mining. Step 3 has no logical foundation since a stock’s price depends on how many shares the company has outstanding. If a firm were to double the number of shares, each share would be worth half as much. There is no reason why a Dow stock with more shares outstanding (and a lower price per share) should be a better investment than a Dow stock with fewer shares outstanding (and a higher price per share). Berkshire Hathaway (which is not in the Dow) has very few shares outstanding and consequently sells for a mind-boggling price of nearly $200,000 per share. Yet it has been a great investment.

What about step 4? Why, after selecting the five stocks with the lowest prices (as if a low price is good), would we cross out the stock with the lowest price? Why indeed.

And steps 5 and 6? Why invest twice as much money in the next lowest priced stock as in the other three stocks? We all know the answer. Because it worked historically. Period.

Shortly after the Gardners launched the Foolish Four Strategy, two skeptical finance professors tested it using data from the years 1949–1972, just prior to the period that had been data-mined by the Gardners. It didn’t work. The professors also retested the strategy during the years that were data-mined by the Gardners, but with a clever twist. Instead of choosing the portfolio on the first trading day in January, they implemented the strategy on the first trading day of July. If the strategy has any merit, it shouldn’t be sensitive to the starting month. But, of course, it was.

In 1997, only one year after the introduction of the Foolish Four, the Gardners tweaked their system and renamed it the UV4. Their explanation confirms their data mining: “Why the switch? History shows that the UV4 has actually done better than the old Foolish Four.” It is hardly surprising that a data-mined strategy doesn’t do as well outside the years used to concoct the theory. The Gardners admitted as much when they stopped recommending both the Foolish Four and UV4 strategies in 2000.

The Foolish Four strategy was indeed foolish.


In the 1980s, an investment advisory firm with the distinguished name Hume & Associates produced The Superinvestor Files, which were advertised nationally as sophisticated strategies that ordinary investors could use to reap extraordinary profits. Subscribers were mailed monthly pamphlets, each about fifty pages long and stylishly printed on thick paper, for $25 each plus $2.50 for shipping and handling.

In retrospect, it should have been obvious that if these strategies were as profitable as advertised, the company could have made more money by using the strategies than by selling pamphlets. However, gullible and greedy investors overlooked the obvious and, instead, hoped that the secret to becoming a millionaire could be purchased for $25, plus $2.50 for shipping and handling.

One Superinvestor strategy was based on the gold-silver ratio (GSR), which is the ratio of the price of an ounce of gold to the price of an ounce of silver. In 1985, the average price of gold was $317.26 and the average price of silver was $5.88, so the GSR was 317.26/5.88 = 54, which meant that an ounce of gold cost the same as 54 ounces of silver.

In 1986 Hume wrote:

The [GSR] has fluctuated widely just in the past seven or eight years, dipping as low as 19-to-1 in 1980 and soaring as high as 52-to-1 in 1982 and 55-to-1 in 1985. But, as you can also clearly see, it has always—ALWAYS—returned to the range between 34-to-1 and 38-to-1.

Figure 4-2 confirms that the GSR fluctuated around the range of 34 to 38 during the years 1970 through 1985.

FIGURE 4-2. The GSR 1970–1985


The GSR strategy is to sell gold and buy silver when the GSR is unusually high and to do the opposite when the GSR is unusually low. Using futures contracts to make these trades creates the potential for astonishing profits.

There is no logical reason why an ounce of gold should cost the same as 36 ounces of silver. As it turned out, after the GSR went above 38 in 1983, it did not come back until twenty-eight years later, in 2011. Futures contracts multiply losses as well as gains, and a 1983 bet on the GSR would have been disastrous.

Figure 4-3 shows that the years when the GSR hovered around 34 to 36 were a temporary fluke, not the basis for a super strategy.

Modern computers can ransack large databases looking for more subtle and complex patterns, but the problem is the same. If there is no underlying reason for the discovered pattern, there is no reason for deviations from the pattern to self-correct.

The moral is simple: Don’t bet the bank on historical patterns that have no logical basis.

FIGURE 4-3. The GSR 1970–2010



Computerized trading systems remove all human judgment. The computers are programmed to track stock prices, other economic and noneconomic data, and news stories, looking for patterns that precede stock price movements. For example, the computers might notice that after the number of stocks going down in price during the preceding 140 seconds exceeds the number going up by more than 8 percentage points, the S&P 500 usually rises.

The computer files this indicator away and waits. When this signal appears again, the computer moves fast, buying thousands of shares in a few seconds and then selling these shares seconds later. Done over and over, day after day, a profit of a few pennies (or even a fraction of a penny) in a few seconds on thousands of shares can add up to real money. The technology magazine Wired gushed that these automated systems are “more efficient, faster, and smarter than any human.”

True, these programs process data faster than any human, but they are no smarter than the humans who write the code that guides the computers. If a human tells a computer to look for potentially profitable patterns—no matter whether the discovered pattern makes sense—and to buy or sell when the pattern reappears, the computer will do so—whether it makes sense or not. Indeed, some of the human brains behind the computers boast that they don’t understand why their computers decide to trade. After all, their computers are smarter than them, right? Instead of bragging, they should be praying.

On May 6, 2010, the U.S. stock market was hit by what has come to be known as a “flash crash.” Investors that day were nervous about the Greek debt crisis, and an anxious mutual fund manager tried to hedge his portfolio by selling $4.1 billion in S&P 500 futures contracts. The idea was that if the market dropped, the losses on this fund’s stock portfolio would be offset by profits on its futures contracts. This seemingly prudent transaction somehow triggered the computers. The computers bought many of the futures contracts the fund was selling, then sold them seconds later. Futures prices started falling and the computers were provoked into a trading frenzy as they bought and sold futures contracts among themselves, like a hot potato being tossed from hand to hand.

Nobody knows exactly what unleashed the computers. Remember, even the people behind the computers don’t understand why their computers trade. In one fifteen-second interval, the computers traded 27,000 contracts among themselves, half the total trading volume, and ended up with a net purchase of only 200 contracts at the end of this fifteen-second madness. The trading frenzy spread to the regular stock market, and the flood of sell orders overwhelmed potential buyers. The Dow Jones Industrial Average fell nearly 600 points (more than 5 percent) in five minutes. Market prices went haywire, yet the computers kept trading. Procter & Gamble (P&G), a rock-solid blue-chip company, dropped 37 percent in less than four minutes. Some computers paid more than $100,000 a share for Apple, Hewlett-Packard, and Sotheby’s. Others sold Accenture and other major stocks for less than a penny a share. The computers had no common sense. They blindly bought and sold because that’s what their algorithms told them to do.

The madness ended when a built-in safeguard in the futures market suspended all trading for five seconds. Incredibly, this five-second time-out was enough to persuade the computers to stop their frenzied trading. Fifteen minutes later, markets were back to normal and the temporary 600-point drop in the Dow was just a nightmarish memory.

There have been other flash crashes since and there will most likely be more in the future. Oddly enough, Procter & Gamble was hit again on August 30, 2013, on the New York Stock Exchange (NYSE) with a mini flash crash, so called because nothing special happened to other stocks on the NYSE and nothing special happened to P&G stock on other exchanges.

Inexplicably, nearly 200 trades on the NYSE, involving a total of about 250,000 shares of P&G stock, occurred within a one-second interval, triggering a 5 percent drop in price, from $77.50 to $73.61, and then a recovery less than a minute later. One lucky person happened to be in the right place at the right time and bought 65,000 shares for a quick $155,000 profit. Why did it happen? No one knows. Remember, humans aren’t as smart as computers.

Fortunately, value investors are inoculated from the perils of technical analysis, since value investors do not try to predict stock prices. Value investors buy a stock because it is an inexpensive money machine, generating bountiful cash, it is hoped, over many, many years.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.