Merely coincident relationships exist in the database, and they are unlikely to hold in a future trade. Using this website, it is easy to find pairs with statistically significant correlation coefficients, but only you can judge whether there is enough of a real economic relationship to be worth considering.

To assist you with this task, you have two filters at your disposal: absolute correlation level and length of history.

A correlation coefficient of 80% is more significant than a 70% correlation, but it is equally significant to a -80% correlation. A correlation of 9% on 100 years of data is equally significant to a 90% correlation on 1 year of data. The longer the historical period studied the more reliable the statistical result will be.

The statistical significance level, say 99%, means that the likelihood that this result could occur purely by chance is 1%. Thus a higher statistical significance level reflects a lower likelihood of a false positive. A search of 1 million pairs filtered to the 90% significance level should have apprxomately 100,000 purely coincidental pairs and at the 99% significance level there should be approximately 10,000 purely coincidental pairs. This is still alot of combinations to consider, but it helps.

A larger correlation coefficient, in absolute value, is more significant, and a correlation coefficient based on more history is more significant. For these reasons, please only consider pairs with at least 75% correlatin over 15 years of history.

If you wish to trade a market which hasn't been at least 15 years or an instrument which has changed its character. You may be forced to use a shorter time periods, but be advised that false positives will be more frequent.

Margin trading involves interest charges and risks, including the potential to lose more than deposited or the need to deposit additional collateral in a falling market. Options are not appropriate for all investors.

For more information, please refer to our Risk Disclosure.