|
What makes NorthStar’s modeling unique?
The core concept underlying NorthStar’s predictive scoring system is the physics principle of a self-organizing phenomenon. The analysis is dynamic; the model measures each variable daily and assigns a proportionate weighting based on how well each variable is correlating with the current market. Over the past 25 years, analysis of an extensive list of broad concepts has been empirically tested for validation. Proved concepts are optimized to derive the most value from the analysis.
Our integration of technical and fundamental analysis has no imposed bias for making bets on the direction of the market. There are no built-in assumptions that historical growth is good, that historical price momentum is good, or that any particular pricing pattern is desirable. Rather, the system calculates daily how the evolving market is correlating historical variables with future price-performance for a series of future time-periods (varying across a range from two days to nine months). Based on the strength and endurance of these correlations, composite predictive scoring values are computed for qualified stocks and merged to produce a final ranking of the domain. Stocks with the highest score become the targets for new opening long positions, and stocks with the lowest scores are the new short candidates. Each time-period modeled uses time-dependent correlations, not an average correlation. Our methodology provides serious insight about how the market works in both trends and turbulence across a wide spectrum of stocks capitalizations. Our database of fundamental data, which extends back over 20 years, is updated weekly from Form10-K and Form 10-Q filings.
NorthStar Methodology
NorthStar’s core model is a pattern matching system that makes extensive use of both pricing and fundamental data. It is based on a very broad principle of self-organization; modeling the market as a driven system with positive feedback. There are numerous driving forces operating in the market including the dynamics of business cycles, tax codes, customer needs, government regulations, and innovative technology. The influence of evolving market forces drives the fundamental data in the SEC filings which is coupled with the strong feedback of the pricing action as the market votes on what all the fundamental data means. The interaction of the fundamental data with the pricing action describes how the market self-organizes. This self-organization is manifest in the distribution functions of market behavior being pushed far away from random behavior.
|
Tracking how markets consistently deviate from random behavior is the objective of the modeling. The ranking system is comprised of 73 scoring models each analyzing market action from different perspectives. Some of the scoring models analyze multi-dimensional distributions of changes in price, trading volume, and volatility. Others use coordinate systems for comparing relative strength (cumulative pricing changes) versus endurance (how often relative out-performance occurs). Other measurements model stocks on fundamental growth metrics (7 metrics including growth of sales, earnings, assets, working capital), quality metrics (10 metrics including margin and turnover metrics), and value metrics (6 metrics including per-share metrics relative to price). Company fundamental values are updated as data from 10-Q and 10-K filings become available. Many of the scoring systems within the oscillator model perform comparisons both at intra-industry and cross-market levels.
The oscillator platform is neither a momentum nor a contrarian-based system. Rather, the methodology seeks to identify characteristic ways the market self-organizes, defining self-organization as the correlation between the historical parameters and the future pricing action. The oscillator scores the equities based on this self-organization, rank ordering all the equities that qualify for tracking. It ranks all equities on daily, weekly, monthly, quarterly, and yearly future price changes. Trading models are then built that are free to weight these multiple future time periods in a manner appropriate to the expected hold-times and the cost of trading.
|
|
