By Ruggiero, Murray A., Jr.
Futures (Cedar Falls, IA) , Vol. 40, No. 10
Intermarket analysis uses known relationships between markets to develop trading rules. It has proven a useful approach for a number of markets, including the S&P 500, Treasuries, Eurodollars, gold, crude oil and more. Numerous magazine articles and books have examined the effectiveness of this technique. An early source of note was technician John Murphy's first book, "Intermarket Technical Analysis" (John Wiley and Sons, 1991).
The book uses the 1987 crash to lay out an intermarket hypothesis. Even by the mid-1990s, these concepts still were new. At the time, there was little public verification they worked. Many institutional traders used the concepts, but mechanical rules generally were not available pubicly.
While you can make market predictions using other techniques, intermarket analysis is one of the few that offers a strong basis for making predictions based on solid fundamental reasoning.
Correlation is not prediction
When most traders get started with intermarket analysis, they begin by looking for correlated markets, or markets whose prices tend to move in tandem. Correlation is important, but it's not as critical as many think. Even when one perfectly correlated market leads another, it can be misleading.
Consider the same Treasury bond series with one shifted five days backward (see "Perfect lead?" below). The correlation of these two markets is not consistent and in fact even dips into negative territory at times. This is why we can't become infatuated with high-correlated markets. Here, we have what should be the perfect lead market--the market itself--and a simple correlation measure lacks precision. What we are interested in is auto-correlation. Auto-correlation is a measure of similarity of two waveforms as a function of a time lag applied to one of them. The success of a particular trading system can indicate sufficient auto-correlation for an intermarket relationship to be profitable.
Before we move to developing mechanical trading systems around intermarket relationships, let's look at the basis of a simple moving average. The lag of a simple moving average is half of the size of the look-back; a 10-day moving average has a lag of five bars. This concept is important and underlies a concept called intermarket divergence. Here's how intermarket divergence works.
For positively correlated markets:
* If the lead market is in an uptrend, and the traded market a downtrend, buy.
* If the lead market is in a downtrend, and the traded market an uptrend, sell.
For negatively correlated markets:
* If the lead market is in an uptrend, and the traded market in an uptrend, sell.
* If the lead market is in a downtrend, and the traded market is in a downtrend, buy.
Various tools can be used to define an up or downtrend. The moving average works well. One way to define a trend would be the sign of price relative to a selected moving average length. The traded market and lead market can have different length moving averages. The code for a system based on this concept is shown in "Developing in divergence" (right).
As an exercise, let's apply our model to the Treasury data set. Because we are looking for divergence, we only will make trades near market turning points based on our perfect data. When optimizing, we found the best set of parameters is 1, 1, 5 and 10. This is a 2.5-day lag for the Treasury bonds and a five-day lag for our shifted five-day forward series. Our Close-Average(Close, X,0) is close minus a closing price centered X/2 days ago.
In our fantasy world, we would make $1,884,968.75. More important, though, is that we still would lose on a few trades. The winning percent is 92%. This demonstrates that the theory is not perfect. The reasons are auto-correlation within the bond series and the changing lengths of market moves.
Unfortunately, we don't have a perfect intermarket signal with a consistent lead. …