The main obstacle in modeling an equity or commodity market is its
shear complexity. Market movements reflect the sentiments of millions
of participants. These sentiments can change rapidly and dramatically.
No single equation can react to all these conditions – one
that successfully predicts market conditions in one week may be
completely inaccurate the next.
The Path Integral development team dealt with inherent complexity
by building a system that combines thousands of subsidiary models.
Each of the independent models utilizes one or more estimating methods
derived from statistical or engineering practices, including but
not limited to regressions and neural nets. Six master algorithms
evaluate these models, deciding which are likely to perform best
given current conditions. It then develops a consensus decision from
suggestions made by the subsidiary models and produces a signal to
long, short, or stay out.
|
System Performance