Econocasts

Friday, February 24, 2012

2012.02.24 Weekly DJIA $ Gold $ Silver Predictions
















































































At this rate, this project will turn into an academic exercise to see when the models refuse to converge. The DJIA has now broken all Z-score records since 1896!

Visit econocasts.net for free sample trading model downloads. 

2 comments:

Oliver said...

Hello Paulo,

Do you still believe the DJIA model is correct considering the extreme Z value?

Ollie

Paolo said...

Hi Oliver,

I think the model is off on prediction of peaks and troughs by at least 5 months. This model was off by four months predicting the last peak in the DJIA, however, it failed to predict the subsequent recovery to 13,000 - and had the DJIA much lower by 2012. Econocasts business cycle chronicles: 2011.06.19 DJIA Long $ Short Term You can see what happened here: Econocasts business cycle chronicles: Cycle Models DJIA Long & Short Term Predictions

Here the model predicted a high in October of 2011, Econocasts business cycle chronicles: 2011.11.04 Weekly DJIA Long $ Short Term but it lagged five months. While the model does adapt - it has been lagging at least 4-6 months in terms of predicting subsequent peaks and troughs peaks over the past two years.

This iteration predicted a peak on the New Year, and it is now lagging by almost 4 months. Econocasts business cycle chronicles: 2012.02.03 Monthly Long Term DJIA $ Gold $ Silver The predicted peak using current data is still stuck on the New Year. If it turns out in a few months that the DJIA does not peak for another few months, then I'd say the model is approaching useless.

On the positive side, it also suggests that monetary policy has altered the magnitudes of both the periods and phases of previously relatively stable long-term business cycles. This might be of academic interest to students of business cycles. The next iteration of the model, if there is one, might have to incorporate some proxies for monetary policy that might be useful in predicting period and phase alterations.

I was having some success on this aspect of model design but the massive increase of the Fed balance sheet was so anomalous that it had no past pattern matches whatsoever, and was thus dropped as a model factor.