Chart Patterns Without Statistics is Naive Technical Analysis

I am saying the obvious but I know some will deny it: Anyone who makes bets in a game of chance without having a measure of the expectation is a naive gambler. This is what many traders do when they rely on chart patterns without having a reliable measure of the expectation. Nobody should ever risk any money on a chart pattern, confirmed or not, without having gone through the exercise of quantifying the chances of success.

There is currently a double top formed in Russell 2000 Index. Is the double top a good indicator of a trend reversal when formed at the top of an extended rally?  Below is a daily chart of IWM, the popular ETF that tracks the index:

The double top formation is indicated on the chart. What can we say about this pattern? Does it indicate a trend reversal?

To start with, this may be just a random formation. But if we must say something about this pattern and what it indicates, if anything at all, we can only do that in the context of probability and statistics.

One way of analyzing the potential of patterns is by analyzing price action after thousands of similar formations. One of the fundamental problems of classical chart analysis is that patterns can be reversal or continuation formations. Therefore, for a chartist to profit in the longer-term, a measure of the expectation of a specific formation is necessary and it must be stable. This further requires obtaining a distribution of returns of a specific formation. However, the distribution may not even exist or it may not be stationary. It turns out that most of those who use chart patterns to trade, especially retail traders who take advantage of free brokerage tools, are involved in a game of chance unless they can quantify the probabilities and related expectation.

If we assume that someone could go through the exercise of analyzing thousands of similar formations, then the expectation E (amount won per trade on the average) may be measured as follows:

E[Chart pattern] = w × avgW – (1-w) × avgL                (1)

where w is the win rate, avgW is the average winning trade and avgL is the average losing trade (in absolute value). If the expectation is sufficiently large, then one could profit from such formation but only in the longer -term, assuming the distribution from which the sample was taken exists and it is stationary.

I hope the reader understand how difficult it is to base trading decisions on chart patterns unless the statistics are know. Yet, the large number of naive users of such formations is staggering. One quick look in the blogosphere will convince any skeptic.

Note that there are some very experienced traders who have a good “feel” of the market and use these chart formations without proper quantifications. These are rare exceptions and it is difficult to have a measure of their win rate. It is also difficult to know whether any success actually depends on these patterns or on experience with these markets.

As a final note, there have been several academic studies in the past on the effectiveness of chart patterns with conflicting conclusions. Besides the fundamental problem of coming up with an algorithm for indentifying these patterns, there are also differences in the data samples, the markets considered, etc. It is difficult to know whether the results are plagued by selection and data-snooping bias.

Disclosure: no relevant positions.
Charting program: Amibroker

This entry was posted in Technical Analysis and tagged , , . Bookmark the permalink.