The Technical Analysis Adventure: Progress Report
A few postings ago, I confessed that I did not fully accept the orthodoxy that technical analysis fails, and I promised a report on what I found after further research and thinking. Here are three major things I have found so far: that technical analysis should be tied to behavioral hypothesis about investor behavior, that technical analysis needs to be combined with risk control and money management techniques, and that some markets may be more amenable to technical analyisis than others.
• Good technical analysis is tied to behavioral hypotheses about investors or traders
There is no doubt that some technical analysis is highly dubious. Simplistic manuals on TA tend to reduce TA to pattern recognition, suggesting that certain patterns tend to be followed by certain other patterns. It is easy to look at the middle of a chart and listen to the narrative story of what happens to a price after it has entered a triangle, pennant, flag, head-and-shoulders, etc.. What is difficult is to see how one would react to these same pattern truncated at the right hand side of a chart, and have to guess the complete picture. Is this a double top? or is it the first two bumps of a head-and-shoulders pattern? How you act when you haven’t yet seen the whole picture makes a big difference to the outcome. These types of pattern-recognition explanations smack of data mining or worse, “searching on the dependent variable,” where the dependent variable turns out to be some outcome one can already see on the chart. Once you know the outcome, it is easy to tell if one has a double top or a head-and-shoulders pattern; it is also too late to profit by it.
One of the first insights presented as I researched is that there is a difference between a price pattern and a price signal. Price patterns are primarily descriptions of a market’s current tendency and may effectively correspond to the market’s “state of mind”. Price signals are events or actions that tell an investor or trader to do something: buy, sell, cover, exit. What events signal what actions are what makes each trader’s technical recipe unique - some may look for breakouts from established patterns, others look for moving averages crossing one another - but signals should be clear and unambiguous, and set out in a list of trading rules to maintain a trader’s discipline.
It is generally not productive to try to figure out what market pattern is developing and make investment or trading decisions based on where you think the pattern will go (with the possible exception of established trends, if you think it is are still early in the trend). This is where technical analysis novices tend to err. Rather, the important moments are where existing patterns break down, because this signals a change in the balance of power between those who think a security is overpriced and those who think it is underpriced. The opportunities in technical analysis come not so much from discovering what “the market” is thinking, and more from discovering when the market has changed it’s mind. Often, when markets enter a new mood, new patterns emerge, and the process of elimination can improve a trader’s chance of entering the right position until the market changes again.
• Technical Analysis needs to be combined with risk control or money management
When TA practitioners analyze charts and explain what transpired, they often sound as if the price patterns that emerged were the only ones which could possibly have appeared. This obscures the fact that the sequence of patterns following each other is more probabilistic than deterministic, and it also makes skeptics more likely to conclude that chartists are all charlatans. In the probabilistic world, once a wedge pattern breaks, prices usually trend in the same direction they trended before they entered the wedge pattern. This is a tendency, not a rule, and this probabilistic approach is what enables the pattern trader to enter a positive sum game. Protective stops allow positions to have a maximum loss (assuming relatively low slippage or gaps), but allow winning streaks to continue. This way, the cost of being wrong is small, but the benefits from correctly understanding a tendency can be much larger - so much so, that Curtis Faith, one of the original Turtle Traders, points out that maybe 30% of one’s trades can account for nearly 100% of one’s profits.
The key here is that traders who use TA must be careful to exercise good money management skills. It is essentially the same application of the Kelly Criterion used in other quantitative applications. By sizing positions appropriately conservatively, one reduces the chance that any one or two investments can wipe your capital base out. Technical signals do not usually tell you how to size your positions - they merely tell you when to enter or exit.
• Some markets may be more tractable with technical analysis than others
Technical analysis detractors point out that technical analysis should not work if markets for investment are efficient. “Weak form” efficiency states that current prices reflect all the information contained in past price (and volume) data, but that extra digging into fundamental information from company management and financial statements might provide value (the semi-strong formulation states that even this extra information cannot reveal any useful information for generating superior returns).
But not all markets are necessarily efficient. If we believe the CAPM family of models, one must hold the entire market portfolio to have an efficient investment. Virtually anything else incorporates additional firm-specific risks and is therefore not efficient. Even indices, if they do not include all investable assets worldwide and in the right proportions, have some inefficiency embedded in them. Illiquid and smaller, non-analyst followed securities may also have pockets of inefficiency in their prices. If market efficiency destroys the ability of technical analysis to work, there are plenty of assets and portfolios that have inefficiencies to them, and technical analysis may well apply to them.
Note that the type of inefficiency in these examples is marginally different from the information inefficiency quality that is usually used to discredit technical analysis. Some readers have argued to me that one type of efficiency implies the other; although I do not fully understand the connection (see an earlier post on “types of market efficiency”). Still, the weak form efficiency argument is relatively easy to discredit on empirical grounds: if markets are truly efficient, asset prices should not exihibit auto-correlation; in fact they do, and many models include a significant momentum factor in predicting returns. Neither of these should happen in informationally efficient markets.
Indices, currencies, commodities, and - to a certain extent - fixed income, are possibly more amenable to technical analysis than company stocks. I hypothesize this because companies are more complex entities than commodity markets or (paradoxically) entire economies. Companies have managements that to some extent seek to manage the stock price, and have (compared to economies or entire commodity markets) some power to affect the company’s own efficiency of profit generation. This is not to say that TA cannot work on individual company stocks, but merely that there are more competing factors that need to be taken into account.
Indeed, one possibility here is that TA can help explain a component of the non-systematic risk of individual companies. This would suggest that stocks that are highly correlated with the market would not be strong candidates for TA (except to the extent that a market index as a whole is amenable to it), but that stocks with lower correlations may react better to TA. Note that high correlation to the market is related to but not the same as stocks with a high market beta, because a stock can have a high beta and still be less correlated than a stock with a lower beta. One issue with lower correlation stocks is that they may be smaller capitalization and possibly lower liquidity stocks. This means that although TA might apply to smaller stocks with lower correlations to the market, the lower liquidity may make the risk control component of a technical investing strategy unviable.
• Good technical analysis is tied to behavioral hypotheses about investors or traders
There is no doubt that some technical analysis is highly dubious. Simplistic manuals on TA tend to reduce TA to pattern recognition, suggesting that certain patterns tend to be followed by certain other patterns. It is easy to look at the middle of a chart and listen to the narrative story of what happens to a price after it has entered a triangle, pennant, flag, head-and-shoulders, etc.. What is difficult is to see how one would react to these same pattern truncated at the right hand side of a chart, and have to guess the complete picture. Is this a double top? or is it the first two bumps of a head-and-shoulders pattern? How you act when you haven’t yet seen the whole picture makes a big difference to the outcome. These types of pattern-recognition explanations smack of data mining or worse, “searching on the dependent variable,” where the dependent variable turns out to be some outcome one can already see on the chart. Once you know the outcome, it is easy to tell if one has a double top or a head-and-shoulders pattern; it is also too late to profit by it.
One of the first insights presented as I researched is that there is a difference between a price pattern and a price signal. Price patterns are primarily descriptions of a market’s current tendency and may effectively correspond to the market’s “state of mind”. Price signals are events or actions that tell an investor or trader to do something: buy, sell, cover, exit. What events signal what actions are what makes each trader’s technical recipe unique - some may look for breakouts from established patterns, others look for moving averages crossing one another - but signals should be clear and unambiguous, and set out in a list of trading rules to maintain a trader’s discipline.
It is generally not productive to try to figure out what market pattern is developing and make investment or trading decisions based on where you think the pattern will go (with the possible exception of established trends, if you think it is are still early in the trend). This is where technical analysis novices tend to err. Rather, the important moments are where existing patterns break down, because this signals a change in the balance of power between those who think a security is overpriced and those who think it is underpriced. The opportunities in technical analysis come not so much from discovering what “the market” is thinking, and more from discovering when the market has changed it’s mind. Often, when markets enter a new mood, new patterns emerge, and the process of elimination can improve a trader’s chance of entering the right position until the market changes again.
• Technical Analysis needs to be combined with risk control or money management
When TA practitioners analyze charts and explain what transpired, they often sound as if the price patterns that emerged were the only ones which could possibly have appeared. This obscures the fact that the sequence of patterns following each other is more probabilistic than deterministic, and it also makes skeptics more likely to conclude that chartists are all charlatans. In the probabilistic world, once a wedge pattern breaks, prices usually trend in the same direction they trended before they entered the wedge pattern. This is a tendency, not a rule, and this probabilistic approach is what enables the pattern trader to enter a positive sum game. Protective stops allow positions to have a maximum loss (assuming relatively low slippage or gaps), but allow winning streaks to continue. This way, the cost of being wrong is small, but the benefits from correctly understanding a tendency can be much larger - so much so, that Curtis Faith, one of the original Turtle Traders, points out that maybe 30% of one’s trades can account for nearly 100% of one’s profits.
The key here is that traders who use TA must be careful to exercise good money management skills. It is essentially the same application of the Kelly Criterion used in other quantitative applications. By sizing positions appropriately conservatively, one reduces the chance that any one or two investments can wipe your capital base out. Technical signals do not usually tell you how to size your positions - they merely tell you when to enter or exit.
• Some markets may be more tractable with technical analysis than others
Technical analysis detractors point out that technical analysis should not work if markets for investment are efficient. “Weak form” efficiency states that current prices reflect all the information contained in past price (and volume) data, but that extra digging into fundamental information from company management and financial statements might provide value (the semi-strong formulation states that even this extra information cannot reveal any useful information for generating superior returns).
But not all markets are necessarily efficient. If we believe the CAPM family of models, one must hold the entire market portfolio to have an efficient investment. Virtually anything else incorporates additional firm-specific risks and is therefore not efficient. Even indices, if they do not include all investable assets worldwide and in the right proportions, have some inefficiency embedded in them. Illiquid and smaller, non-analyst followed securities may also have pockets of inefficiency in their prices. If market efficiency destroys the ability of technical analysis to work, there are plenty of assets and portfolios that have inefficiencies to them, and technical analysis may well apply to them.
Note that the type of inefficiency in these examples is marginally different from the information inefficiency quality that is usually used to discredit technical analysis. Some readers have argued to me that one type of efficiency implies the other; although I do not fully understand the connection (see an earlier post on “types of market efficiency”). Still, the weak form efficiency argument is relatively easy to discredit on empirical grounds: if markets are truly efficient, asset prices should not exihibit auto-correlation; in fact they do, and many models include a significant momentum factor in predicting returns. Neither of these should happen in informationally efficient markets.
Indices, currencies, commodities, and - to a certain extent - fixed income, are possibly more amenable to technical analysis than company stocks. I hypothesize this because companies are more complex entities than commodity markets or (paradoxically) entire economies. Companies have managements that to some extent seek to manage the stock price, and have (compared to economies or entire commodity markets) some power to affect the company’s own efficiency of profit generation. This is not to say that TA cannot work on individual company stocks, but merely that there are more competing factors that need to be taken into account.
Indeed, one possibility here is that TA can help explain a component of the non-systematic risk of individual companies. This would suggest that stocks that are highly correlated with the market would not be strong candidates for TA (except to the extent that a market index as a whole is amenable to it), but that stocks with lower correlations may react better to TA. Note that high correlation to the market is related to but not the same as stocks with a high market beta, because a stock can have a high beta and still be less correlated than a stock with a lower beta. One issue with lower correlation stocks is that they may be smaller capitalization and possibly lower liquidity stocks. This means that although TA might apply to smaller stocks with lower correlations to the market, the lower liquidity may make the risk control component of a technical investing strategy unviable.