As mentioned early on, the human eye is an incredibly unreliable tool for finding cycles because we humans tend to find patterns anywhere we look for them, even in noise. Our minds do this by accepting data points near a model (that seem to define a pattern) and reject anything that doesn’t.
This phenomenon is pervasive, as anyone who has eyeballed a moving average and concluded that it follows a price trend nicely can attest to.
It can be quite jarring, then, when the moving average is run as a computerized trading system which takes each cross of the moving average precisely, rather than disregarding the very real, significant violations that the eye discounts.
The only way around this is to approach cycle validation logically and mathematically, which brings us back to the use of highly specialized software. As mentioned, this lies beyond the scope of this course. If you’d like to study the topic in depth, we recommend reading Schwagger and Mogey.
For your reference, they outline the eight-step process of cycle identification and validation as follows:
1) Select your data
2) Visually inspect it
3) Transform the data into log form (this accomplishes initial detrending)
4) Smooth the data
5) Find possible cycles
6) Complete detrending using the difference from a moving average
7) Test cycles for their statistical significance and dominance
8) And Combine and project cycles into the future
Each of these steps is absolutely essential and involves more issues than novice (and even some veteran) analysts ever consider relevant.