Stock Charts Fail Forecast Test in Complete S&P Miss (Update3)

By Michael Tsang and Eric Martin

May 4 (Bloomberg) -- John Bollinger, inventor of the “Bollinger bands” system of predicting stock movements with price charts, says technical analysisworks.

“I don’t know what people are saying when they say somehow indicators have broken down,” Bollinger, president of Bollinger Capital Management, said in a telephone interview from Manhattan Beach, California. “It’s like somehow saying streetlights don’t work anymore. As long as people obey them, streetlights work.”

Ever since the Standard & Poor’s 500 Index peaked in October 2007, six of eight strategies -- which are supposed to make money whether stocks rise or fall -- failed, according to back-testing data compiled by Bloomberg. As the bear market erased $11 trillion from the value of U.S. equities, buy and sell signals from those six technical indicators produced losses of as much as 49 percent, the data show.

“Technical analysis on its own as a discipline does not work,” said Diane Garnick, the New York-based investment strategist at Invesco Ltd., which oversees $348 billion. Using it in isolation is “the fastest way to lose money,” she said.

Of the eight strategies, stochastics, Bollinger bands, relative strength, commodity channels, parabolic systems and the Williams %R indicator generated buy and sell signals that resulted in losses between the S&P 500’s peak of 1,565.15 on Oct. 9, 2007, and its March 9 trough, the data show. They did worse as the index then rallied 30 percent through last week.

No Help

The systems failed to protect investors last year, when the S&P 500 had its biggest decline since 1937, as price swings reached a record, according toWilliam Stone, chief investment strategist at PNC Financial Services Group Inc.’s wealth management unit, which oversees $96 billion in Philadelphia.

Bloomberg’s back-testing model purchases the S&P 500 when an indicator signals a “buy” and holds it until a “sell” is generated. At that time, the index is sold and a short position established and kept until a “buy” is triggered.

The S&P 500 added 3.4 percent to 907.24 today, erasing its decline for 2009, after home sales beat forecasts and manufacturing in China expanded.

Bollinger bands are designed to alert investors when a security rises too high or falls too low by comparing its price to the average level over the past 20 days. If the stock gains or drops enough from the average -- two standard deviations -- a turnaround may be at hand, Bollinger’s system says. Moves of two standard deviations are defined as occurring 5 percent of the time or less in a statistical model.

Financial Stocks

When used to determine when to buy or bet against S&P 500 financial stocks, the technique produced a gain through mid- September of $28,588 on a $100,000 investment, when banks retreated 37 percent, data compiled by Bloomberg show.

Profits evaporated in less than a month and turned into a loss of $64,388 as the strategy failed to trigger any sell signals during the rest of the bear market, when banks and brokerages plummeted 72 percent.

John Bollinger says that methodology is too simple and his bands should be used in conjunction with data on trading volume to create “set-ups” and “confirmations” for investment decisions. His fund, which aims to profit in any environment, made money sometimes during the bear market. He declined to comment on specific returns.

Stochastics predicts a security’s movement based on how close its price is to the highest or lowest levels. Stochastics would have left anyone who started with $100,000 at the October 2007 peak with $75,881 by March 9, a 24 percent loss, according to data compiled by Bloomberg.

False Buys

The trades were undone by the buy signals, which on eight occasions suggested that the S&P 500 had fallen too far, too fast, based on data compiled by Bloomberg that exclude trading costs and unexpected price fluctuations at the moment of the trade. Stochastics told traders to buy on Oct. 6 last year, three weeks after Lehman Brothers Holdings Inc.’s bankruptcy. The S&P 500 lost 14 percent in the following month.

Burton Malkiel, whose 1973 investment text “A Random Walk Down Wall Street” argued that price movements aren’t predictable, says chart-based investing worsens returns.

“People who think they are going to make excess profits with technical analysis are kidding themselves,” Malkiel said in a telephone interview from Princeton, New Jersey. “Most of the people who say this is pretty good have some ax to grind.”

Traders shouldn’t use charts without other technical data, and choosing the most appropriate ones can both mitigate losses and produce gains, said Katie Townshend Stockton, chief market technician at Greenwich, Connecticut-based MKM Partners LLC.

Cut Losses

Each of the eight strategies would have cut losses for investors benchmarkedto the S&P 500. The directional movement indicator and the moving average convergence/divergence indicator flashed signals that made money even during the worst financial crisis since the Great Depression.

Directional movement is a theory developed by J. Welles Wilder in 1978 that measures how far a security moves from an average price range calculated from second to second. The system is designed to gauge who is more eager to trade a security, buyers or sellers.

The so-called DMI generated a gain of 24 percent. The moving average convergence/divergence indicator, which bases trades on the difference between 12- and 26-day moving averages, provided profits of $25,896 from a $100,000 investment.

In both strategies, the signals that directed traders to sell and then short the S&P 500 made more money than their buy signals lost. Short sellers typically borrow shares from a brokerage and then sell them on a bet they will be able to repurchase the stock at a lower price.

Value Investors

Returns also exceeded those of many who follow the principles of value investing, the practice popularized by Warren Buffett, the billionaire chairman of Omaha, Nebraska- based Berkshire Hathaway Inc. He searches for the cheapest companies relative to earnings or assets.

Buffett told shareholders at Berkshire’s annual meeting on May 2 that “many” U.S. stocks have fallen to levels where they are cheap relative to their intrinsic value. He said the housing market is the biggest drag on the economy and remains “very hard” to forecast.

Bill Miller, whose Legg Mason Value Trust beat the S&P 500 for a record 15 straight years through 2005, produced a loss of 72 percent, including dividends, during the bear market.

David Dreman was fired this year by Deutsche Bank AG’s asset management unit after his flagship $2.62 billion DWS Dreman High Return Equity Fund lost65 percent. Both managers piled into stocks such as Freddie Mac andAmerican International Group Inc., misjudging the severity of the financial meltdown.

‘A Little Misleading’

“It’s a little misleading to be looking at any indicator in a vacuum,” said MKM’s Stockton. “It’s a matter of knowing which ones to combine and knowing what environment you’re in.”

None of the technical indicators produced a bigger return than S&P 500’s 30 percent rally from its 12-year low on March 9.

Five sent signals that have resulted in losses of between 1.9 percent and 8.3 percent during the eight-week climb, which added $2.29 trillion to the value of U.S. equities, data compiled by Bloomberg show. The Williams %R indicator, which calculates the difference between a security’s closing price and its highest price and then compares the result with its average over 14 days, handed traders who heeded its buy and sell signals an $8,286 loss on $100,000 invested at the trough.

Only one technical indicator, moving average convergence/divergence, made money for investors during both the bear market and the subsequent rebound.

“Some of the indicators might work at given times, but that’s not where I put my odds,” said PNC’s Stone, a certified market technician who learned to read charts on the trading floor of Salomon Brothers Inc. in the early 1990s. “You zigged when you should have zagged.”

Ranked Performance of Eight Technical IndicatorsIndicator                              10/09/2007 - 3/09/2009Relative Strength Index                                -49.0%Williams %R                                            -41.7%Commodity Channel Index                                -38.7%Parabolic Systems                                      -36.6%Bollinger Bands                                        -31.5%Stochastics                                            -24.1%Directional Movement Indicator                         +24.0%Moving Average Convergence/Divergence                  +25.9%S&P 500                                                -56.8%--------Indicator                               3/09/2009 - 5/01/2009Commodity Channel Index                                 -8.3%Williams %R                                             -8.3%Bollinger Bands                                         -6.6%Stochastics                                             -3.3%Directional Movement Indicator                          -1.9%Moving Average Convergence/Divergence                   +7.8%Parabolic Systems                                       +8.2%Relative Strength Index                                +21.8%S&P 500                                                +29.7%--------Indicator                              10/09/2007 - 5/01/2009Williams %R                                            -43.1%Commodity Channel Index                                -40.3%Parabolic Systems                                      -34.3%Relative Strength Index                                -34.1%Bollinger Bands                                        -22.2%Stochastics                                            -21.8%Directional Movement Indicator                          +9.0%Moving Average Convergence/Divergence                  +24.6%S&P 500                                                -43.9%

To contact the reporters on this story: Michael Tsang in New York atmtsang1@bloomberg.net; Eric Martin in New York atemartin21@bloomberg.net.