What Is The Value Of Technical Analysis?

A recent post from the excellent Mathematical Investor blog questions whether the use of chart patterns and technical analysis truly offers value in financial markets.  The authors point out how easy it is to manipulate information to look significant (commonly encountered when an "analog" to the current time period is found in a previous historical period).  Indeed, it's possible to find historical analogs to any market behavior simply because the search space over the course of financial history is so large.  This is classic overfitting: the similarities of today and the past are likely to be chance artifacts.

As one astute market participant noted to me, one has to be suspicious that other disciplines do not make use of chart patterns and indicators of historical time series.  If, for example, a weather forecaster were to note that today's warm temperature is a breakout from the recent range of temperatures and therefore we should see temperatures trending higher through the week, this would not be a credible forecast.  Nor would we take seriously a weather forecast that looked for configurations of cloud patterns.

Although the validity of technical patterns is often questionable (How often do we see valid backtests of assertions made on technical grounds?), it is their poor reliability that I find particularly problematic.  It is not unusual to find two technicians look at the same chart and arrive at radically different conclusions based upon the lookback period considered and the definition of the pattern.  One might see one wave count in a given market; another will arrive at a different count.  Both will entertain "alternate counts" that lead to radically different conclusions.  Can you imagine radiologists arriving at wildly different interpretations of imaging scans?  The lack of reliability would make it difficult to develop any kind of valid surgical intervention.

All that being said, I do see empirical work out there that links past returns to future ones.  Very often, these studies find value and momentum effects (circumstances in which past returns lead to reversals or continuation) that are tested for economic as well as statistical significance.  I have also seen traders firmly define patterns that "set up" in intraday markets and test them out for skews in forward returns, creating successful "playbooks" that guide their trading.  This study of market intraday momentum recently came to my attention as an example of more rigorous implementation of price patterns as potential predictors.  I also observed a daytrader this past week rigorously test a pattern of behavior in the VWAP of stocks that led to short-term momentum.  

Does technical analysis have merit?  I would argue yes, but more as a source of hypotheses than as a source of conclusions.  We can frame market behavior in terms of patterns, but it is important that these patterns be defined objectively and tested properly before they merit the investment of hard-earned dollars.

Further Reading:  A Hard Look at the Performance of Market Forecasters

.