Q Spectrum module
I have conducted a lot of research regarding the new Q Spectrum module. A whole week of a heavy job, scientific research. From one side, I see that this spectrum is much more better for revealing cycles that forecast. As a target, the forecast ability of this cycle is used while classical spectrum searches fitness, the cycle that fits the price. Speaking mathematically, classical spectrum deals with INTERPOLATION while Q spectrum deals with EXTRAPOLATION (i.e. how this cycle forecasts). Somehow this difference is not accented in classical cyclical analysis. I talked many times about this subject: classical cyclical analysis solves other tasks, we cannot blindly apply these techniques as they are for the stock market, we have to combine them with some kind of confirmation.
OK, now I am trying to find the optimal steps to apply this spectrum for trading. The new possibilities lead to new problems. This is the other side - new unknown (for me, at least) problems, searching some missing link.
As it is now, the most promising approach looks quite interesting; it is explained below. I call it double verification method. Definitely this is not a final picture, this is just a beginning of a huge job to be done.
As it is now, it works this way:
We download the price history and set LBC 1-5 years back. Though it depends on the cycle period you are looking for. Typically I try to find the cycles with the period 25-250 calendar days; in this case LBC 1-5 years back is OK. If you are looking for long term cycles, it is better to use more price history on your testing interval.
In this example I will show how to reveal Annual cycle. I have downloaded DD (Du Pont), don't forget to remove splits and set LBC to 2009, i.e. 7 years back:
Run Q Spectrum. As you see, it sets backtesting mode automatically:
You see, it shows strong Annual cycle, click on it:
and look what it shows: before LBC walk forward efficiency (WFE) of this cycle = 76.7% i.e. for 23 years it provided good forecast versus not good forecast for 7 years.
But it also conducts walk forward analysis after LBC and tells us that 5 years after LBC it has worked good versus 1 year not good. I mean these fragments of info record:
In other words, we have conducted a double verification for this cycle.
The same manner I've checked other cycles including inverted cycles, like this 65.79 days cycle that works not bad: