Image source: http://www.forex.in.rs/wp-content/uploads/2011/11/forex_stocks_correlation_2.png
So what is the correlation of financial instruments? Correlation is a quantity reflecting the degree of similarity of the graphs of the two instruments. In fact, the introduction of a correlation is an attempt to express the degree of similarity of graphs with just one value, called the correlation coefficient. The value of this coefficient varies from -1 to 1. Where 1 denotes the maximum degree of similarity, when growth on one graph is always accompanied by a similar increase in strength on the other, 0 is the absence of similarity, and -1 is the reverse similarity, when growth on one is accompanied by a proportional fall on the second.
Since the correlation, in fact, reflects only the degree of similarity of the graphs, it is not at all necessary that in the presence of a high correlation coefficient, there will be some real interconnection between the trading instruments. It is quite possible that two graphs for absolutely random reasons will be similar to each other. But if the number of graph points is large enough, i.e. The correlation is statistically reliable, then the probability of accidental coincidence becomes negligible. In this case, we can talk about the existence of a relationship between financial instruments, which ensures the similarity of their schedules.
It is the ability of correlation to measure the interrelation between financial instruments makes it an extremely useful tool for pair trading. Recall what is pair trading. This is a multidirectional trade of interrelated tools, where profit is extracted by playing to eliminate imbalances that periodically arise between these tools. The key word here is "interconnected", since the success of the strategy will depend directly on how strong and real the relationship between the pair of instruments is. But how to discover interconnected tools among the hundreds of options available on various exchanges? And here the correlation coefficient comes to the rescue. It is enough to sort through all possible pairwise combinations of trading instruments, calculate the correlation coefficients for them and select those from which this indicator will be quite high (for example, more than 0.8).
In order not to do such complex calculations yourself, you can use online services to calculate the correlation. One of the most convenient free services is Correlation of pairs on the site megatrader.org. This service displays pairwise correlation of tools in the form of a correlation table, allows you to specify tool lists and time frames. He is also interested in the fact that, in addition to the correlation itself, it is possible to immediately calculate the weight coefficients for the instruments that make up the pair, and also to plot the spread of the pair. To do this, just click on any correlation value in the table, and the page with a spread graph and automatically calculated weights will open on the site. By the way, these coefficients can be adjusted manually and see how this will affect the spread schedule.
Of course, among the selected pairs with high correlation there will most likely be some percentage of pairs with a false dependence. There are two ways to combat this. First, during selection, it is necessary to pay attention to the fact that both instruments are from the same sector of the economy. In this case, the probability of accidental coincidence is significantly reduced. Secondly, it is worthwhile to trade not with one pair, but with a portfolio of pairs. Then, even if some pairs turn out to be random, their losses will be compensated by the rest of the pairs, and the total return on the portfolio will remain positive.
