For a while now I have intended to develop a low level cyclic analysis of trend data that is quick to perform, universal and robust. The goal is to quickly identify and extract major local and overall cycles in data trends from underlying asset price performance. Using strong cycles, then guestimating if the reported fundamentals supports a continuation or possible modification of the cycles.
It is obvious to me that the cyclic nature of news, sentiment, spin, deception and outright price forces (i.e. trading) is not magic. As a rule, I ignore all main stream news and annecdotal/opinion broadcasts, observe sentiment and rely 95% on trusted data correlated to stated results.
Applying my latest project to the USDx (US Dollar Index) the results are encouraging.
(note: analysis below was optimised for most recent USDx performance only – being the 2008 to 2012 period).
Cycles are observed in correlations also. Too many people (incl me previously) assume a continuous and constant correlation with USDx, which is not the case at all. I note that statistics does not have a big role to play in the above, but for the statistically minded, Rsquared is better than 65%.