Econocasts

Saturday, January 16, 2016

2016.01.15 Gold Cycle Model Chart

2016.01.15 Gold Cycle Model Chart


The gold cycle model continues to phase shift forwards, also dragging the timing of the predicted bottom forward. Nevertheless, the model continues to suggest that the $1000 level will likely not be breached during the phase extension, if it continues. The previous iteration of the model is shown below and on the blog here.

2015.112.04 Gold Cycle Model Chart

6 comments:

Diego said...

Thanks Paolo. I "feel" we are getting close now. ;)

Paolo said...

Hi Diego, I think you are probably right. I'm glad you are getting a "feel" for the model.

Diego said...

Hi Paolo, among many other things, I figured that nobody commenting this Post about Gold for over a week was probably a good indicator! ;)
Greetings and thanks.

Paolo said...

Well Diego, I'll take some blame for that since I may have been remiss in explaining how to interpret model results. Some readers expect perfection, but that comes from ignorance of structural model theory. Suffering comes from expectation.

AM said...

Paolo, first of all thanks for posting pretty forecast pictures!
Your gold model forecasted increase in gold prices for last two months of 2015 - instead prices crashed over 100$/Oz. That is a loss of 10,000$ per one gold futures contract. This is not the first time we see deviations from the model like this. How useful is this model for traders? I understand the academic interest, sure. For traders a much more accurate model is required, dont you think?
Thanks again for sharing.

Paolo said...

All good points, AM. Because these cycle models have a lot of error, both in magnitude as well as time, they give an imperfect view of the future. They don't contain any information except for previous price and time. In the case of gold, the phase extension due to model error has pushed to well over a year. Completely useless for day trading or weekly trading or monthly trading.

What the models do provide is a probability of price movement based on the difference between the actual and predicted curves, captured as a Z-score. In the case of gold, some investors were calling for $600-$800/oz prices, which according to the model would have been an eight-sigma event. My interpretation of the gold model prediction is that while the phase extension might continue, gold is clearly heading for a major rise but the timing of that rise may continue to be pushed into the future, but the probability of a significant drop in current price is quite small, but not zero.

For me, the models have additional utility because all of the neural net models I use have the Z-score as an input, and the information derived from the Z-score is weighed heavily by the neural model. As to how well a neural net model performs, my best neural net model, which happens to be the VIX which I use for day to weekly trading because of the large price swings, has a tested out-of-sample success rate of 68%, which is the best I have ever been able to achieve. In contrast, the neural net gold model has a success rate of 62%, which is also very reasonable, but to those who do not use quantitative models a 62% success rate may appear to be a failure.

So, taken together, I consider the cycle models as a sort of first fuzzy look into the future of prices, but the information provided is only one data point a frequent trader might consider in the calculus of when to buy and when to sell.