The theoretical basis for the work

Price curves in the markets are a classic example of an auto-correlated time series. The market can be seen as a set of parallel brains integrating prices and financial information which alter buying/selling which in turn alters the price levels and generates new financial information. The "waves" of herding behavior amongst traders result in non-linear statistical signatures in the price-time series which can be exploited for fun and profit, because on the whole, the parallel market-brain system generates the data upon which it acts - a classic recursive function.

Because fear and greed are the relatively constant human emotions which drive trading, some patterns tend to repeat themselves.   The time series is analyzed as a
multifractal in both time and price dimensions based on the work of Benoit Mandelbrot. Proprietary routines are used to find matches in patterns of price/time by comparing the non-linear parameters obtained from the scaling  analyses. A comparison of the non-linear parameters to the original time series is analyzed for statistical validity  and the valid patterns are used to reconstruct and project time series behavior into the future. The current VIX models are implemented in a Support Vector Machine (SVM)/neural network.

In answer to the question of "If you're so good why do you need to sell information?" the answer is simple. Monetizing the information we use for our own investments is an efficient way of gleaning additional capital for investment.