This cannot possibly work because "any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes" and Danielsson's corollary "A risk model breaks down when used for regulatory purposes".
The very moment any model - regardless how cleverly designed - is published and started to be used to make money on a significant scale, peolple will start gaming the system. This will necessarily destroy any information carrying correlation. Why? Because, as it is impossible to generally directly predict economic success except after the fact any more than it is possible generally to predict whether a given program will halt or not w/o running it, you have to use proxies. These proxies are always easier to manipulate on purpose than they are by influencing them as a side effect to a different purpose.
A good example is the number of papers published as a measure for scientific excellence and predictor of future academic success. A good indicator - but only as long a nobody counts them and uses it to direct money flows. The moment you do that, researches will see publication as an end in itself, and not as a byproduct of successful research and the measure will lose most of its predictive power.
In economics, it's even worse, as money made on speculation spends exactly the same as money earned by building value: Charts of stock prices would be an excellent predictor as market prices in theory should aggregate all available information - but only as long nobody uses them to estimate value or predict future prices. With chart analysis used by many market actors and companies being allowed to manipulate their own stock price (vie buy-backs, option programs, etc.) you get so many artificial feedback effects into the system that they dominate the system's dynamics and the actual signal gets lost in the noise.