FAQ
Why is Level E
going to perform better than traditional fund
management?
Level E aims to incorporate both expert knowledge and statistically
optimal methods within a probabilistic and AI inspired framework. By
using these formal frameworks to integrate many thousands of potential
features, MAYA™ has an advantage over larger organizations where the
integration of differing opinions can be less formally optimal, and
therefore less reliable.
What can Level E
offer?
Level E can provide predictive signals indicating probable market
behavior, allowing more informed investment decisions. In addition,
Level E can provide a complete investment portfolio taking into account
diversification requirements and additional client constraints.
How does Level E
limit over exposure during multiple stock failures?
Once signals are produced and expected returns are
calculated,
diversification of model-stock pairs helps to ensure against exposure
to multiple failures. Clusters of similarly behaving model-stock pairs
are dynamically created, and so diversification is achieved by creating
a portfolio from multiple clusters. Furthermore, exposure levels are
fine-tuned at the execution stage, to incorporate further
diversification constraints.
What is a stock feature, and
how are they used?
A stock feature can be thought of as a “shape” or “motif” within the
stock time-series data. These shapes can be flexible along certain
predefined dimensions, such as height of maximum price increase within
a time window, or changes in the stock price trend over time. Often
these features are similar to those used by human traders, but within
the MAYA™ framework they can be optimized, and formally integrated as
inputs to a principled model.
What is a
probabilistic model, and how are they different to
traditional methods?
A probabilistic model is a mathematically formal way
of dealing with
uncertainty in data, and coping with underlying or hidden factors
influencing the data. The crucial difference between probabilistic
models and traditional methods is that a probabilistic model will
maintain probability distributions over the search space, whereas
traditional models generally operate according to sets of rules. Using
distributions allows formal integration of defined prior beliefs into
the model which ensure that the distribution in the model is altered by
exactly the confidence defined in the prior distributions. Without a
probabilistic model, any rules are effectively defining a prior
probability distribution, but in a non-formal manner, which means that
it is often unclear how to optimize such models.
Formal probabilistic models have the ability to marginalize over hidden
states, meaning that by computing expectations these models can
properly deal with unknown variables and hidden market dependencies.
MAYA™ aims to formally integrate a very large selection of stock
features as inputs to multiple probabilistic models. In this manner,
expert knowledge and rules can be incorporated into features, and
integrated in a mathematically optimal fashion.
What markets can be
covered?
Level E's technology is not limited to any
particular market, and
currently Level E has successfully employed investment solutions
working in the UK's developed market (FTSE100 constituents), and also
in Brazil's emerging market (IBOVESPA constituents). Level E's
technology is currently adaptable to different market situations such
as limitations on short selling, or investment period restrictions.
Work is currently in progress to tailor to ever more specific
constraints such as those imposed by Shari'a compliant funds.
As well as straight equities investment, Level E's technology is
capable of managing other investment solutions such as CFD trading or
spread betting.
|
