Theory

Wed, March 14, 2012

Volatility Research sole purpose has been to "crack the VIX volatility code." We are pleasantly surprised that our VXX daily forecasts of VXX next-day close have done so well this year.

Our scorecard so far this year is better than we expected, even though our algorithm was optimized to predict VXX gains, not losses, late last year (Sep-Dec) in a period of much higher volatility (see charts 4,5,6) than now:

31 right
9 wrong
9 no forecast
78 / 100 score (31 / 40)

THEORY* We are unable to explain why our forecasts this year are far more right than wrong. We know we should be in murky waters using historical data to forecast this year's VXX price movements. Our only possible explanation now is this:

"VXX price movements are caused by the confluence of many variables occurring randomly, of which their sets are not mutually coexistent and are dissimilar over time; VXX price movements today over 4 consecutive days that are similar to historical price movements over 4 consecutive days, i.e. they have the same qPath as previously, will have, with reasonable certainty, the same 5th day price direction following this year's qPath as the same qPath had previously — either up or down — even though the set of variables affecting volatility and VXX ensuing price today is distinctly different from historical variables."

Its one theory we can't prove, other than "it seems to be working."

We are not aware of any other forecasts that have done this well. We hope that everyone interested in VIX volatility can benefit from our forecasts — and they're FREE! Just bookmark this page for our latest.

* CORRECTION: A previous version of our Theory implied we were using only historical data from last year. We are using historical data from each previous day through last year.