Making the Case for the Trade One of the Most Classic Forms of Analysis Is Finding Past Market Action Similar to Today's and Using What Happened in the Past to Forecast the Future. However, Such Analog Analysis Often Is Approached Incorrectly. Here's the Scientific Way
Ruggiero, Murray A., Jr., Futures (Cedar Falls, IA)
A-nal-o-gy (e nal'e je), n., pl. -gies. agreement or resemblance in certain aspects, similarities between dissimilar things.
The use of analog analysis in trading is the comparison of charts or sections of charts in an effort to find agreement or similarity so that what transpired on one chart can be used to forecast what will happen on the other chart. It is a trading methodology that attracts attention every few years. Usually, the interest is in the area of chart patterns and seasonal analogs.
Developing a seasonal analog for a market is a relatively simple process. It involves taking a snapshot of the same seasonal time window of data and comparing it year to year for a number of years seeking an identifiable seasonal pattern. For example, look at the charts for prices of corn from June of one year to the following June and sort the price action using different patterns that the market traces out over this period. The goal of this exercise is to find similar patterns that are predictive of future price movement. In our analysis of the seasonal analog for corn from June of one year to June of the following year, we might want to attempt to forecast price action in corn for the time period from July to November.
There is another type of analog study that does not use seasonality as its basis. This type of analog looks at chart patterns over a period of time and tries to find similar patterns that are predictable. These analog studies can be long-term comparisons, trying to locate the correlations between the stock market crash of 1929 and what happened to the Nasdaq index in 2000 and 2001, for example. Analog studies can even be used to compare and correlate different markets. For example, an analog study could be made comparing the Nasdaq chart from 1996 to 2000, natural gas from 2000 to 2001 and the move in coffee from the bottom in 1993 to the top in 1994. These charts all share the distinction of being examples of parabolic bull markets.
Parabolic moves are an interesting type of market move because they always retrace to at least the point where the move went parabolic. This has been the case in each of these three market moves. The 1996 low in coffee was a retracement to the point where the move went parabolic in 1994. Natural gas retraced to a point that was below the 2000 low in early 2002. The Nasdaq also retraced to a price level much lower than the point where the market went parabolic at about 1700.
After this type of move, the peak prices often remain all-time high levels for many years to come. Natural gas had a major retracement in late 2002 but the early 2003 peak was much lower. It looks like natural gas will continue to consolidate at lower levels after its parabolic peak in 2001 (see "Gas bubbles," right).
Analog studies can also be applied to charting patterns such as triangles, head-and-shoulder patterns, pennants and other repetitive chart formations. The point is that similar chart formations should produce similar future price action, which means that they can be used as a predictive tool. Analog studies can also be made on patterns of two to five bars. Larry Williams' Oops and Smash patterns are examples of analog pattern studies that are predictive.
PROBLEMS WITH ANALOGS
Analog analysis is a powerful tool. However, there is a problem with its implementation because it is hard to backtest and identify analogs using mechanical means except for simple price bar patterns that are a few bars in length or seasonal tendencies that are relatively simple. As an example, consider the bars in "Gas bubbles." It would be difficult to write the code to identify the analog patterns in these charts and develop a gauge that would tell how predictive the patterns actually are.
One of the most difficult things to do in programming systems is to write code that identifies chart patterns such as head-and-shoulders formations, triangles or pennants. …