The database collects daily sentiment numbers (market consensus) to analyze market behavior in a â€œbottom upâ€ fashion in order to determine the state of market opinion and predict the likely hood of market movements based on contrary opinion and historical price action based on archived behavioral patterns. By looking at sentiment in a historical and behavioral manner of market movement (dating back to 1964), various mathematical concepts such as Random Walk and Bernoulli are used to qualify and measure the path and statistical characteristics of behavioral patterns. The measurements are calculated and stored within archives that are recalled and provide performance statistics to measure the market's tendency to be either overbought or oversold. This systematic approach is technically based on a proprietary model that evaluates price over time combined with market consensus to create a system signal; both must be met before taking a position. We look at the patterns of consensus numbers over a certain time frame, not just the latest number. Consensus is the instrument that provides the main filter to determine the strength of price breakouts and whether the system should initiate a position against these breakouts. When a market does break out of a trading range, this system will attempt to participate by entering on a re-test of the breakout. A comprehensive database of string patterns and analysis of trade action has been created and is used to gauge statistical probabilities of future direction.
Our edge lies in that we have a large enough historical sample to gauge how markets behave (based on a particular behavior pattern) over a history of varying market and economic conditions. This vast historical dataset allows us to identify what side of the market is wrong and will need to correct and place our bet based on statistical outcomes that have occurred over a 50 plus year timeframe, in varying market conditions. We illustrate this by applying the Random Walk principle and superimpose actual behavior with what the Random Walk would look like in order to see what happens over multiple years and where we think the "street" is likely to be making the same mistakes as in the past. Measurements are calculated and stored by archives that we can recall based on specific aggregations of behavior. The system exploits these distortions by taking directional positions in the major liquid commodity, currency and bond markets.
By statistically spreading this approach across multiple markets, the system's performance is designed to follow the overall statistical performance of the historical behavior using artificial intelligence. This Machine Learning system perpetually acquires and archives data on a daily basis. So in this sense the system is perpetually growing and getting smarter as it acquires more history that lends to the statistical forecasts it produces.