As a child did you lie on your back in the grass looking up at the clouds? Did you see lions, houses, faces and other shapes? If so, perhaps you would be good at charting.
In Part 3 of our series of articles on finding value in the stock market we consider charting and technical analysis. Charting consists of the careful study of past price movements and perhaps past trading volumes in order to make forecast about future prices. I had a friend who would buy large rolls of charting paper and scroll them across his work bench. He would meticulously mark price movements and then overlay them with triangles of various shapes and sizes. His aim was to detect hidden triangular patterns because he believed that they would be repeated allowing him to forecast future movements of the price.
I doubt many people do it these days in such a laborious way. Eyeballing yards and yards of chart paper has been replaced by sophisticated computer packages sniffing out past patterns. This is called technical analysis.
In the first part of the article I want to sketch the ideas of three of the most influential chartists, Charles Dow, Ralph Elliott and William Gann.
Dow Theory
When charting, perhaps people don't search for lions and elephants, but they do look for other interesting shapes with suggestive names such as double bottoms and head and shoulders. The original school of charting was based on 255 Wall Street Journal articles written by Charles H. Dow (1851-1902), the founder and first editor of the Wall Street Journal and co-founder of Dow Jones and Company with Edward Jones and Charles Bergstresser.
After his death William P. Hamilton, Dow's understudy and the fourth editor of the Wall Street Journal, organized the material and presented it in a more coherent form. According to John Murphy in his book Technical Analysis of the Financial Markets, the basic tenets of Dow Theory are:
Averages discount everything.
The market has 3 trends.
Major trends have 3 phases.
The averages must confirm each other.
Volumes must confirm the trend.
A trend is considered to be in effect until there are definite signals showing otherwise.
As an example of a trend, Dow defined an uptrend as a time when successive rallies in a security price close at levels higher than those achieved in previous rallies and when lows occur at levels higher than previous lows.
The three phases of a trend are accumulation, a public participation, and distribution.
The idea that trends must be confirmed by volume is that when price movements are accompanied by higher volume, Dow believed this represented the "true" market view. If many participants are active in a particular security, and the price moves significantly in one direction, Dow maintained that this was the direction in which the market anticipated continued movement. To him, it was a signal that a trend is developing.
Elliott Waves
Another major school of charting is based on the ideas of Ralph Nelson Elliott (1871-1948), a professional accountant. His idea was that market prices unfold in specific patterns which practitioners today call Elliott waves. Elliott published his views of market behavior in a number of books culminating in Nature's Laws - The Secret of the Universe published in 1946.
He was not short on confidence writing that "because man is subject to rhythmical procedure, calculations having to do with his activities can be projected far into the future with a justification and certainty heretofore unattainable."
A major proponent of Elliott wave theory is Robert Prechter who has written many books on Elliott waves and their applicability to forecasting markets. One of his forecasts has been consistent: "One thing I've repeated consistently," he said, "is that the great bear market will take the DJIA at least below 1,000 and likely to below 400." Phew!
A few years ago I shared the platform with Prechter at a conference on finance and investing. After his talk, in my innocence I asked if the earnings of companies played a role in the stock price. He looked at me as if I was from another planet. In the official proceedings of the conference he stated that exogenous forces such as "economic reports, wars and peace treaties, terrorism, elections, corporate earnings, scandals, Fed actions and the movements of other markets" are irrelevant. "None of these classes of events," he continued, "has a leading or coincident relationship to stock price movement."
Another aficionado of Elliott is Richard Swannell who wrote software called Elliottician. I met him back in 2001 and asked if he had made any money from his system. He said that at that point he had not used it with his own money. Perhaps things have changed since then since I notice on his site that he states, "I will personally give US$100,000 to anyone who can prove they have a more accurate market forecasting tool than my Refined Elliott Trader (RET) software. This can include any software program or any technical indicator, whatsoever..."
Starting with Elliott himself and the quote by him above, then moving on to Prechter, and finally to Swannell, it seems that self-assurance is at stellar levels in the world of Elliott waves.
Gann Charts
The next main character in the Parthenon of the most influential charting proponents is W. D. Gann (1878-1955). His idea is that the market will make highs at integer multiples of the all-time low and the timing will be quantifiable from the value of this low. It also applies to a lessor extent to major lows and highs.
Gann believed that the ideal balance between time and price exists when prices rise or fall at a 45 degree angle relative to the time axis. This is also called a 1 x 1 angle since prices rise one price unit for each time unit.
Gann Angles are drawn between a significant bottom and top (or vice versa) at various angles. the 1 x 1 trendline was considered the most important by Gann and signified a bull market if prices are above the trendline or a bear market if below. Gann felt that a 1 x 1 trendline provides major support during an up-trend and when the trendline is broken, it signifies a major reversal in the trend. Gann identified nine significant angles, with the 1 x 1 being the most important. They are:
1 x 8 - 82.5 degrees
1 x 4 - 75 degrees
1 x 3 - 71.25 degrees
1 x 2 - 63.75 degrees
1 x 1 - 45 degrees
2 x 1 - 26.25 degrees
3 x 1 - 18.75 degrees
4 x 1 - 15 degrees
8 x 1 - 7.5 degrees
Gann observed that each of the angles can provide support and resistance depending on the trend. For example, during an up-trend the 1 x 1 angle tends to provide major support. A major reversal is signalled when prices fall below the 1 x 1 angled trendline. According to Gann, prices should then be expected to fall to the next trendline (i.e., the 2 x 1 angle). In other words, as one angle is penetrated, expect prices to move and consolidate at the next angle.
The following is a typical Gann chart or Gann fan. It would be interpreted as the index "bouncing" off the 2 x 1 and 1 x 2 lines.
One major stumbling block with this method is that it depends on the units that you use since it is not always practical to give the 1x1 line a value of 1 point of price for each day. For example, if the S&P500 is trading around 1,500 it is not sensible to consider time units of 1,500 days. Gann said that another scale should be used but did not give any rules as to how to do this.
Although you can buy computer packages to implement the above methods, as they were described by their developers the forecasts could be achieved using pencil and paper with a bit of care. The next step in the story is to use computer tools to search for patterns and features as part of making forecasts. This is called technical analysis and the following are brief descriptions of some of the main technical analysis tools
Moving Average Operators
This is an average, possibly weighted, of prices over some time period. Because it is an average, it is filters out high frequency data which could be interpreted as noise and so is taken to represent more accurately the "true" trend of the price. They are referred to as low pass filters.
Long-term moving averages are slower to respond than short-term moving averages. Hence a typical trading rule would be to buy when:
the 50-day moving average is slowly rising, and
the 15-day moving average crosses from below the 50-day moving average.
Another way of describing these two requirements is that there was an overall slight upward trend which has recently accelerated.
Channel Breakout Operators
A channel is the price range between the lowest price and the highest price over a period of time. If the price moves above the high end of the channel, then it is a signal to buy. Conversely, if it breaks out in the lower direction, it is a signal to short the stock.
Stochastic Operators
These operators measure the relative position in a channel over a specified number of days. Because it is a relative position, it removes the trend in the price within a moving channel. It is the opposite of moving average operators since it acts as a high pass filter.
Bollinger Bands
This approach was developed by John Bollinger in the early 1980s as a band whose position and width varied according to price movements. The band consisted of three curves. The middle curve is a measure of the intermediate-term trend, usually a simple moving average, that serves as the base for the upper and lower edges. The interval between the upper and lower edges and the middle edge is determined by volatility, typically the standard deviation of the same data that were used for the average.
A typical example of a Bollinger band is:
Middle Curve: 20-day simple moving average
Upper Edge: Middle Curve + 2 * 20-day standard deviation
Lower Edge: Middle Curve - 2 * 20-day standard deviation
It is an adaptive extension of the stochastic operators mentioned above.
Some traders treat the band as a type of "break out operator" and buy when the price breaks above the upper edge and sell when it drops below the lower edge. Others take the opposite view and buy when prices touch the lower edge and exit when price touches the moving average in the center of the bands.
Geometrical Patterns
There are an unlimited range of classes of geometrical patterns. These include the famous head and shoulders pattern with the suggestive rule that you should short the stock when the neckline support is broken.
These should be enough examples to show you that the range of charting and technical analysis systems and rules is enormous.
Two question remain. Why should we expect any of the rules to generate excess profits where by excess profits I mean trading profits above what could be expected by random guessing after transaction costs? And, even if we expect any of them to work, is there any evidence that they actually generate excess profits in a practical way?
The Elliott Wave group tend to state that their approach is based on real physical laws of nature related to the rhythms of nature. In this regard, Robert Prechter likes to emphasize the role of herding. He wrote, "Under conditions of uncertainty, people instinctively impulsively herd together and make decisions in ways that have little to do with conscious, rational thought." It is this herding, according to Prechter, that gets expressed in predictable patterns of waves.
Others take a more insouciant view about technical analysis. For example, in their book Technical Analysis of Stock Trends, Robert D. Edwards and John Magee write, "We can never hope to know why the market behaves as it does … History obviously has repetitive tendencies and that's good enough."
The trouble with the view that "if it works, its good enough for us," is that there is scant consistent evidence that it actually does work. For example, in a recent book Evidence-Based Technical Analysis by David Aronson, the author back tested 6,402 technical analysis rules on the S&P 500 over the period from November 1, 1980 through to July 1, 2005.
In the final chapter of the book, Aronson writes: "No rules with statistically significant returns were found." None, zero, nada.
Looking at the above rules and strategies from the perspective of a mathematician, I can't help seeing a type of "gee whiz" quality about them. The people who discovered them have applied mathematics to what seemed a chaotic situation and are excited about their achievements. This seems to be followed by an attachment to their ideas.
In each case the mathematics is really very elementary but nevertheless generates some attractive charts and images. The trap is that mathematics is incredibly seductive but, for many, also very intimidating. As soon as a few equations or geometric figures are attached to some price charts, many people think that the methods are legitimized and proven.
More Advanced Methods
Even though Aronson tested over 6,000 rules, he just scraped the surface of known technical analysis strategies. For example, there is a whole range of attempts to use neural networks and more sophisticated mathematics such as fractal geometry and chaos theory to understand and forecast markets.
As an example, a paper by Mark Leung, Hazem Daouk and An-Sing Chen in the International Journal of Forecasting titled Forecasting stock indices: a comparison of classification and level estimation models looks at technical analysis forecasting using a range of mathematical techniques including probabilistic neural network methods. They show that from January 1991 through December 1995 their methods on average outperformed the S&P 500 FTSE 100 Nikkei 225 Index by 3.59% per year.
Better than nothing, but not too remarkable given that no mention was made of transaction costs and slippage which is the inability to actually make the transaction at the quoted price. Also there was no discussion of draw down levels and possible excess volatility of the trading portfolio.
Another approach is to use intermarket influences. For example, currently I am an examiner for a PhD thesis in which the candidate uses neural networks to see if information on overseas markets can be used to make successful daily movement forecasts of the All Ordinaries Index in Australia.
At least qualitatively, it is well recognized in today's world of global markets aided by high-speed international financial transactions there is a lot of influence between markets. Here are a few recent headlines in Australian newspapers: “Stocks in the red after Wall St slide”, “Stocks follow Wall St lower at open”, “Stocks open flat after Wall St fall” and “Stocks open higher after Wall St rebound”.
In fact, it is a cause of some surprise when the Australian market does not follow the US as indicated by the recent headline in Australia: “Stocks surge despite US malaise”. The PhD thesis uses neural networks to show that it is possible to increase the accuracy of forecasts on the Australian market by using information from overseas markets.
As far as anyone can tell, the trading system with the most computational power has been developed by the Prediction Company in Santa Fe. It is a secretive company founded in March 1991 by a group of physicists with the aim of coupling the most advanced mathematics with a powerful array of computers to conquer the world's markets. On their website they claim:
Our technology allows us to build fully automated trading systems which can handle huge amounts of data, react and make decisions based on that data and execute transactions based on those decisions - all in real time. Our science allows us to build accurate and consistent predictive models of markets and the behavior of financial instruments traded in those markets... We have a substantial track record with excellent results.
We don't actually know if these claims are true. Reading the book The Predictors by Thomas Bass describing the years of the Prediction Company up until 1999 certainly gives rise to doubts. For example, describing the state of development in 1995, Bass writes, "Their ambitious project to model the most actively traded stocks on the New York Stock Exchange remains a gleam in their eye." He explains that that at the time they are waiting for a new methodology to come online.
This is such a common story told by technical traders. Everything will be fine and the profits will start to roll in. All that is needed are a few adjustments to the parameters, or the purchase of a new software system from the suppliers which just happens to cost twice as much as the current version. The successful results always seem to be imminent, just around the corner.
Up until the writing of the book by Bass, eight or nine years after the formation of the company, this was the story with the Prediction Company which is now fully owned by UBS Warburg.
Even if you have a trading system that works in principle, there is the problem of human greed and arrogance when it comes to its implementation. Consider the case of the Nobel laureates and other founders of Long Term Capital Management who ended up causing the company to lose billions of dollars. Roger Lowenstein provides a vivid account in When Genius Failed of how some of the most brilliant minds in the financial world dazzled bankers around the world with their reputations and mathematical prowess before crashing to earth caught in their own hubris.
For a period in the mid 1990s I also became caught up in the goal of creating a super trading system. I wrote a large amount of code for trading markets using volatility. My observation was that after periods of high volatility, the volatility settles back to its long-term levels. Similarly, after periods of low volatility it will increase to its long term levels. The idea was to be able to measure accurately the volatility relative to its long term levels and to trade it accordingly using instruments such as option straddles.
The other tool that I was involved with at the time was looking for inconsistencies in foreign exchange option markets. At least this one had a concrete outcome; I wrote a book with Valery Kholodnyi on the topic called Foreign Exchange Option Symmetry.
The Randomness of Markets
Over the short-term, it is very difficult to show that charts formed by equity prices and indices are anything but random. In fact, trillions of dollars of contracts and transactions are based on the assumption that they are random. As an example, the ubiquitous Black and Scholes formula for option pricing that you see in the annual reports of most companies valuing the options awarded to management and staff assumes that the prices follow a random walk.
Specifically, the formula assumes that the returns on a day by day basis follow a normal or Gaussian probability distribution. Most agree these days that these distributions are not normal, that they should have fatter tails. On this topic I recommend Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets by Nassim Nicholas Taleb.
The point is that even if we don't agree on the type of randomness, to a very high degree price performance is extremely close to random and it is only with the use of sophisticated mathematics and powerful computers that they can be distinguished from random.
Methods for generating random graphs throw up all the features of charting and technical analysis described above. A number of studies have been carried out with chartists to see if they could distinguish between actual price charts and randomly generated charts. They could not.
The chart below gives an example of how a simple computer program can be used to build up a chart in systematic, but random, way so that the final result looks very similar to a typical price chart.
The chart is taken from the introduction to Derivatives and Financial Mathematics edited by John Price
Such charts can show all the features of typical price charts but because they are randomly generated, it is not possible to use past data to make any forecasts of future behavior.
Since this is part of a series of articles on finding value in the stock market, it is only fair that I state what I think about charting and technical analysis as a method for finding value. Despite all the marketing hyperbolae I think that it is highly unlikely that anyone working on their computer at home with a retail trading package (or one that they designed themselves) is going to make money in anything like a consistent manner. If you have 1000 people, some will make money during any given year. But over time, most people will probably lose money.
Perhaps if you had access to the forecasts of the Prediction Company, and you had very deep pockets to weather the drawdowns, you might be successful.
My advice if you are tempted to purchase a trading package: ask to see the actual trading records on the tax returns of the person who wrote the code and of the senior personnel in the company.
In the fourth article in this series I am going to look at the valuation methods of Benjamin Graham, the man who is called the Dean of Wall Street. Apart from describing his methods, I will also examine their relevance in today's markets.
In Part 3 of our series of articles on finding value in the stock market we consider charting and technical analysis. Charting consists of the careful study of past price movements and perhaps past trading volumes in order to make forecast about future prices. I had a friend who would buy large rolls of charting paper and scroll them across his work bench. He would meticulously mark price movements and then overlay them with triangles of various shapes and sizes. His aim was to detect hidden triangular patterns because he believed that they would be repeated allowing him to forecast future movements of the price.
I doubt many people do it these days in such a laborious way. Eyeballing yards and yards of chart paper has been replaced by sophisticated computer packages sniffing out past patterns. This is called technical analysis.
In the first part of the article I want to sketch the ideas of three of the most influential chartists, Charles Dow, Ralph Elliott and William Gann.
Dow Theory
When charting, perhaps people don't search for lions and elephants, but they do look for other interesting shapes with suggestive names such as double bottoms and head and shoulders. The original school of charting was based on 255 Wall Street Journal articles written by Charles H. Dow (1851-1902), the founder and first editor of the Wall Street Journal and co-founder of Dow Jones and Company with Edward Jones and Charles Bergstresser.
After his death William P. Hamilton, Dow's understudy and the fourth editor of the Wall Street Journal, organized the material and presented it in a more coherent form. According to John Murphy in his book Technical Analysis of the Financial Markets, the basic tenets of Dow Theory are:
Averages discount everything.
The market has 3 trends.
Major trends have 3 phases.
The averages must confirm each other.
Volumes must confirm the trend.
A trend is considered to be in effect until there are definite signals showing otherwise.
As an example of a trend, Dow defined an uptrend as a time when successive rallies in a security price close at levels higher than those achieved in previous rallies and when lows occur at levels higher than previous lows.
The three phases of a trend are accumulation, a public participation, and distribution.
The idea that trends must be confirmed by volume is that when price movements are accompanied by higher volume, Dow believed this represented the "true" market view. If many participants are active in a particular security, and the price moves significantly in one direction, Dow maintained that this was the direction in which the market anticipated continued movement. To him, it was a signal that a trend is developing.
Elliott Waves
Another major school of charting is based on the ideas of Ralph Nelson Elliott (1871-1948), a professional accountant. His idea was that market prices unfold in specific patterns which practitioners today call Elliott waves. Elliott published his views of market behavior in a number of books culminating in Nature's Laws - The Secret of the Universe published in 1946.
He was not short on confidence writing that "because man is subject to rhythmical procedure, calculations having to do with his activities can be projected far into the future with a justification and certainty heretofore unattainable."
A major proponent of Elliott wave theory is Robert Prechter who has written many books on Elliott waves and their applicability to forecasting markets. One of his forecasts has been consistent: "One thing I've repeated consistently," he said, "is that the great bear market will take the DJIA at least below 1,000 and likely to below 400." Phew!
A few years ago I shared the platform with Prechter at a conference on finance and investing. After his talk, in my innocence I asked if the earnings of companies played a role in the stock price. He looked at me as if I was from another planet. In the official proceedings of the conference he stated that exogenous forces such as "economic reports, wars and peace treaties, terrorism, elections, corporate earnings, scandals, Fed actions and the movements of other markets" are irrelevant. "None of these classes of events," he continued, "has a leading or coincident relationship to stock price movement."
Another aficionado of Elliott is Richard Swannell who wrote software called Elliottician. I met him back in 2001 and asked if he had made any money from his system. He said that at that point he had not used it with his own money. Perhaps things have changed since then since I notice on his site that he states, "I will personally give US$100,000 to anyone who can prove they have a more accurate market forecasting tool than my Refined Elliott Trader (RET) software. This can include any software program or any technical indicator, whatsoever..."
Starting with Elliott himself and the quote by him above, then moving on to Prechter, and finally to Swannell, it seems that self-assurance is at stellar levels in the world of Elliott waves.
Gann Charts
The next main character in the Parthenon of the most influential charting proponents is W. D. Gann (1878-1955). His idea is that the market will make highs at integer multiples of the all-time low and the timing will be quantifiable from the value of this low. It also applies to a lessor extent to major lows and highs.
Gann believed that the ideal balance between time and price exists when prices rise or fall at a 45 degree angle relative to the time axis. This is also called a 1 x 1 angle since prices rise one price unit for each time unit.
Gann Angles are drawn between a significant bottom and top (or vice versa) at various angles. the 1 x 1 trendline was considered the most important by Gann and signified a bull market if prices are above the trendline or a bear market if below. Gann felt that a 1 x 1 trendline provides major support during an up-trend and when the trendline is broken, it signifies a major reversal in the trend. Gann identified nine significant angles, with the 1 x 1 being the most important. They are:
1 x 8 - 82.5 degrees
1 x 4 - 75 degrees
1 x 3 - 71.25 degrees
1 x 2 - 63.75 degrees
1 x 1 - 45 degrees
2 x 1 - 26.25 degrees
3 x 1 - 18.75 degrees
4 x 1 - 15 degrees
8 x 1 - 7.5 degrees
Gann observed that each of the angles can provide support and resistance depending on the trend. For example, during an up-trend the 1 x 1 angle tends to provide major support. A major reversal is signalled when prices fall below the 1 x 1 angled trendline. According to Gann, prices should then be expected to fall to the next trendline (i.e., the 2 x 1 angle). In other words, as one angle is penetrated, expect prices to move and consolidate at the next angle.
The following is a typical Gann chart or Gann fan. It would be interpreted as the index "bouncing" off the 2 x 1 and 1 x 2 lines.
One major stumbling block with this method is that it depends on the units that you use since it is not always practical to give the 1x1 line a value of 1 point of price for each day. For example, if the S&P500 is trading around 1,500 it is not sensible to consider time units of 1,500 days. Gann said that another scale should be used but did not give any rules as to how to do this.
Although you can buy computer packages to implement the above methods, as they were described by their developers the forecasts could be achieved using pencil and paper with a bit of care. The next step in the story is to use computer tools to search for patterns and features as part of making forecasts. This is called technical analysis and the following are brief descriptions of some of the main technical analysis tools
Moving Average Operators
This is an average, possibly weighted, of prices over some time period. Because it is an average, it is filters out high frequency data which could be interpreted as noise and so is taken to represent more accurately the "true" trend of the price. They are referred to as low pass filters.
Long-term moving averages are slower to respond than short-term moving averages. Hence a typical trading rule would be to buy when:
the 50-day moving average is slowly rising, and
the 15-day moving average crosses from below the 50-day moving average.
Another way of describing these two requirements is that there was an overall slight upward trend which has recently accelerated.
Channel Breakout Operators
A channel is the price range between the lowest price and the highest price over a period of time. If the price moves above the high end of the channel, then it is a signal to buy. Conversely, if it breaks out in the lower direction, it is a signal to short the stock.
Stochastic Operators
These operators measure the relative position in a channel over a specified number of days. Because it is a relative position, it removes the trend in the price within a moving channel. It is the opposite of moving average operators since it acts as a high pass filter.
Bollinger Bands
This approach was developed by John Bollinger in the early 1980s as a band whose position and width varied according to price movements. The band consisted of three curves. The middle curve is a measure of the intermediate-term trend, usually a simple moving average, that serves as the base for the upper and lower edges. The interval between the upper and lower edges and the middle edge is determined by volatility, typically the standard deviation of the same data that were used for the average.
A typical example of a Bollinger band is:
Middle Curve: 20-day simple moving average
Upper Edge: Middle Curve + 2 * 20-day standard deviation
Lower Edge: Middle Curve - 2 * 20-day standard deviation
It is an adaptive extension of the stochastic operators mentioned above.
Some traders treat the band as a type of "break out operator" and buy when the price breaks above the upper edge and sell when it drops below the lower edge. Others take the opposite view and buy when prices touch the lower edge and exit when price touches the moving average in the center of the bands.
Geometrical Patterns
There are an unlimited range of classes of geometrical patterns. These include the famous head and shoulders pattern with the suggestive rule that you should short the stock when the neckline support is broken.
These should be enough examples to show you that the range of charting and technical analysis systems and rules is enormous.
Two question remain. Why should we expect any of the rules to generate excess profits where by excess profits I mean trading profits above what could be expected by random guessing after transaction costs? And, even if we expect any of them to work, is there any evidence that they actually generate excess profits in a practical way?
The Elliott Wave group tend to state that their approach is based on real physical laws of nature related to the rhythms of nature. In this regard, Robert Prechter likes to emphasize the role of herding. He wrote, "Under conditions of uncertainty, people instinctively impulsively herd together and make decisions in ways that have little to do with conscious, rational thought." It is this herding, according to Prechter, that gets expressed in predictable patterns of waves.
Others take a more insouciant view about technical analysis. For example, in their book Technical Analysis of Stock Trends, Robert D. Edwards and John Magee write, "We can never hope to know why the market behaves as it does … History obviously has repetitive tendencies and that's good enough."
The trouble with the view that "if it works, its good enough for us," is that there is scant consistent evidence that it actually does work. For example, in a recent book Evidence-Based Technical Analysis by David Aronson, the author back tested 6,402 technical analysis rules on the S&P 500 over the period from November 1, 1980 through to July 1, 2005.
In the final chapter of the book, Aronson writes: "No rules with statistically significant returns were found." None, zero, nada.
Looking at the above rules and strategies from the perspective of a mathematician, I can't help seeing a type of "gee whiz" quality about them. The people who discovered them have applied mathematics to what seemed a chaotic situation and are excited about their achievements. This seems to be followed by an attachment to their ideas.
In each case the mathematics is really very elementary but nevertheless generates some attractive charts and images. The trap is that mathematics is incredibly seductive but, for many, also very intimidating. As soon as a few equations or geometric figures are attached to some price charts, many people think that the methods are legitimized and proven.
More Advanced Methods
Even though Aronson tested over 6,000 rules, he just scraped the surface of known technical analysis strategies. For example, there is a whole range of attempts to use neural networks and more sophisticated mathematics such as fractal geometry and chaos theory to understand and forecast markets.
As an example, a paper by Mark Leung, Hazem Daouk and An-Sing Chen in the International Journal of Forecasting titled Forecasting stock indices: a comparison of classification and level estimation models looks at technical analysis forecasting using a range of mathematical techniques including probabilistic neural network methods. They show that from January 1991 through December 1995 their methods on average outperformed the S&P 500 FTSE 100 Nikkei 225 Index by 3.59% per year.
Better than nothing, but not too remarkable given that no mention was made of transaction costs and slippage which is the inability to actually make the transaction at the quoted price. Also there was no discussion of draw down levels and possible excess volatility of the trading portfolio.
Another approach is to use intermarket influences. For example, currently I am an examiner for a PhD thesis in which the candidate uses neural networks to see if information on overseas markets can be used to make successful daily movement forecasts of the All Ordinaries Index in Australia.
At least qualitatively, it is well recognized in today's world of global markets aided by high-speed international financial transactions there is a lot of influence between markets. Here are a few recent headlines in Australian newspapers: “Stocks in the red after Wall St slide”, “Stocks follow Wall St lower at open”, “Stocks open flat after Wall St fall” and “Stocks open higher after Wall St rebound”.
In fact, it is a cause of some surprise when the Australian market does not follow the US as indicated by the recent headline in Australia: “Stocks surge despite US malaise”. The PhD thesis uses neural networks to show that it is possible to increase the accuracy of forecasts on the Australian market by using information from overseas markets.
As far as anyone can tell, the trading system with the most computational power has been developed by the Prediction Company in Santa Fe. It is a secretive company founded in March 1991 by a group of physicists with the aim of coupling the most advanced mathematics with a powerful array of computers to conquer the world's markets. On their website they claim:
Our technology allows us to build fully automated trading systems which can handle huge amounts of data, react and make decisions based on that data and execute transactions based on those decisions - all in real time. Our science allows us to build accurate and consistent predictive models of markets and the behavior of financial instruments traded in those markets... We have a substantial track record with excellent results.
We don't actually know if these claims are true. Reading the book The Predictors by Thomas Bass describing the years of the Prediction Company up until 1999 certainly gives rise to doubts. For example, describing the state of development in 1995, Bass writes, "Their ambitious project to model the most actively traded stocks on the New York Stock Exchange remains a gleam in their eye." He explains that that at the time they are waiting for a new methodology to come online.
This is such a common story told by technical traders. Everything will be fine and the profits will start to roll in. All that is needed are a few adjustments to the parameters, or the purchase of a new software system from the suppliers which just happens to cost twice as much as the current version. The successful results always seem to be imminent, just around the corner.
Up until the writing of the book by Bass, eight or nine years after the formation of the company, this was the story with the Prediction Company which is now fully owned by UBS Warburg.
Even if you have a trading system that works in principle, there is the problem of human greed and arrogance when it comes to its implementation. Consider the case of the Nobel laureates and other founders of Long Term Capital Management who ended up causing the company to lose billions of dollars. Roger Lowenstein provides a vivid account in When Genius Failed of how some of the most brilliant minds in the financial world dazzled bankers around the world with their reputations and mathematical prowess before crashing to earth caught in their own hubris.
For a period in the mid 1990s I also became caught up in the goal of creating a super trading system. I wrote a large amount of code for trading markets using volatility. My observation was that after periods of high volatility, the volatility settles back to its long-term levels. Similarly, after periods of low volatility it will increase to its long term levels. The idea was to be able to measure accurately the volatility relative to its long term levels and to trade it accordingly using instruments such as option straddles.
The other tool that I was involved with at the time was looking for inconsistencies in foreign exchange option markets. At least this one had a concrete outcome; I wrote a book with Valery Kholodnyi on the topic called Foreign Exchange Option Symmetry.
The Randomness of Markets
Over the short-term, it is very difficult to show that charts formed by equity prices and indices are anything but random. In fact, trillions of dollars of contracts and transactions are based on the assumption that they are random. As an example, the ubiquitous Black and Scholes formula for option pricing that you see in the annual reports of most companies valuing the options awarded to management and staff assumes that the prices follow a random walk.
Specifically, the formula assumes that the returns on a day by day basis follow a normal or Gaussian probability distribution. Most agree these days that these distributions are not normal, that they should have fatter tails. On this topic I recommend Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets by Nassim Nicholas Taleb.
The point is that even if we don't agree on the type of randomness, to a very high degree price performance is extremely close to random and it is only with the use of sophisticated mathematics and powerful computers that they can be distinguished from random.
Methods for generating random graphs throw up all the features of charting and technical analysis described above. A number of studies have been carried out with chartists to see if they could distinguish between actual price charts and randomly generated charts. They could not.
The chart below gives an example of how a simple computer program can be used to build up a chart in systematic, but random, way so that the final result looks very similar to a typical price chart.
The chart is taken from the introduction to Derivatives and Financial Mathematics edited by John Price
Such charts can show all the features of typical price charts but because they are randomly generated, it is not possible to use past data to make any forecasts of future behavior.
Since this is part of a series of articles on finding value in the stock market, it is only fair that I state what I think about charting and technical analysis as a method for finding value. Despite all the marketing hyperbolae I think that it is highly unlikely that anyone working on their computer at home with a retail trading package (or one that they designed themselves) is going to make money in anything like a consistent manner. If you have 1000 people, some will make money during any given year. But over time, most people will probably lose money.
Perhaps if you had access to the forecasts of the Prediction Company, and you had very deep pockets to weather the drawdowns, you might be successful.
My advice if you are tempted to purchase a trading package: ask to see the actual trading records on the tax returns of the person who wrote the code and of the senior personnel in the company.
In the fourth article in this series I am going to look at the valuation methods of Benjamin Graham, the man who is called the Dean of Wall Street. Apart from describing his methods, I will also examine their relevance in today's markets.
