Onyx Financial is a provider of financial market technologies focusing on creating
robust solutions using the latest quantitative methods and technologies.
Algorithmic Trading
Onyx Financial's trading platform interfaces with a number of brokers and provides
real time automated trading. New models and algorithms can easily be added and ensembled
to stay up to date with market conditions and risk profiles. Failover connectivity
or VPS hosting may be enabled to ensure seamless access to the market.
The platform back-tests models against a wide variety trading objectives and walk-forward
scenarios. Onyx Financial has an extensive library of financial data spanning the
past 60 years in the equity and FX markets and 10 years in the commodities markets.
Many of Onyx’s models have been back-tested on over 100,000 bars of historical data.
One of Onyx Financial’s key focuses is on keeping the number of features and parameters
in a model small to ensure that the model is robust and not over-fitted to aged
market conditions. Models also usually combine different time frames and the ensembling
of principles from the fields of Pattern Recognition, Intermarket-Relationships
and Digital Signal Processing (DSP).
Onyx Financial is not currently trading fixed income, credit or volatility based
instruments however this is currently under investigation.
Pattern Recognition and Machine Learning
One of Onyx Financial’s key strengths of is its innovative use of genetic and machine
learning computational techniques to gain the maximum returns from market inefficiencies.
In Onyx Financial’s arsenal of pattern recognition models include –
Support Vector Machines (SVMs)
Gaussian Process Regressions (GPRs)
K-Nearest Neighbours (KNNs)
Artificial Neural Networks (ANNs)
Most of these systems are able to provide a Bayesian measure of confidence allowing
trading to only occur on those patterns with the highest confidence.
Many patterns in financial time series occur over wildly different time frames. Onyx
Financial implements models which draw heavily from the field of Speech Recognition
including Isolated Word Recognition (IWR) to dynamic warp market data feeds to correct
for dilation in the time or volatility domains. This also gives the algorithms a
tendency to be adaptive to market conditions.
Onyx Financial employs genetic algorithms to solve the problem of parameter optimisation.
However, using a reduced number of parameters or robust sets of parameters which
perform well over a wide variety of market conditions is favoured to trying to find
the absolute optimal solution. Additionally non-parametric models which map feature
vectors to higher dimensions are favoured as they require minimal optimisation and
are hence more stable.
Model inputs are measured for effectiveness and carefully chosen using a number of
proprietary metrics. The distribution of these inputs is usually examined and mapped
to a uniform probability distribution to increase the model’s ease of learning.
Furthermore, the number of features is frequently reduced using the Karhunen–Loève
Transform (KLT) or Singular Value Decomposition (SVD). This ensures that the principal
components of market behaviour are identified and isolated before modelling.
Onyx Financial frequently ensembles many models into a single trading algorithm.
This ensures that the maximal amount of inefficiency is extracted from the historic
time series.
Shannon Entropy, Mutual Information, Hidden Markov Models (HMMs), Maximum Entropy
Spectral Analysis (MESA) and Singular Spectrum Analysis (SSA) models are currently
under investigation.
Volatility Modelling
Onyx Financial has a wide variety of techniques for volatility approximation and
forecasting. At the most simple are the EWMA and GARCH models which we use for scaling
of time series data and in our risk management methodology.
In the field of derivatives trading, accurate forecasting of volatility is crucial.
Onyx Financial is able to use its machine learning models to create high fidelity
models which simplify the measurement and prediction of volatility and hence yield
accurate derivative pricing. This may be employed for volatility arbitrage, VIX
trading, or pricing of derivatives on both buy and sell sides of OTC transactions.
Onyx Financial is able to provide a number of derivative pricing models based on
Finite Difference Method (FDM) based on either explicit or implicit calculations.
Risk Management
Risk Management is the most commonly overlooked component of systematic trading.
Onyx Financial solves this challenge by combining a number of simple yet robust
strategies.
Firstly, Onyx Financial is able to deliver experience in Monte Carlo methods for
both Market and Credit Risk Management. Of particular interest are the use of Antithetic
and Low Discrepancy Numbers.
However for the purpose of risk management on real time trading, a volatility adjusted
bootstrap of historic portfolio values can yield a highly robust measure of the
expected draw-down with a given level of confidence. This avoids the need for excessive
calibration or approximation of correlations. In other words this uses the co-dependency
structure of the historical data to model portfolio behaviour while avoiding the
need and hence dangers inherent in calibration. Historical changes in portfolio
value are corrected for volatility before bootstrapping. The success of this method
is reliant on having a reasonable approximation of the current volatility.
Onyx Financial is able to combine many trading strategies optimally to minimise portfolio
drawdown. This is tempered with an understanding that in extreme events the co-dependency
of markets is likely to be very different to normal day-to-day trading. In other
words, in a distressed market, most instruments’ behaviour tends to approach perfect
correlation.
Using Onyx Financial’s systems a new trade is not taken unless it satisfies risk
management criteria.
Technologies
Onyx Financial is able to provide extensive experience in software development. Our
custom solutions are architected and built to fit business needs. These predominately
focus on robust, fast, usable and extensible systems. Our development technologies
of choice are –
Microsoft Visual C# 2010
Microsoft Visual C++ 2010
Microsoft SQL Server 2008 R2
Mathworks MATLAB
Microsoft Excel 2010 (VBA)
Web technologies such as ASP.Net 4.0 are also available on request.
Where appropriate we have the option of performing highly vectorised calculations
on the CPU using Intel’s Maths Kernel Library (MKL) library.
We are currently investigating hardware-dedicated GPU processing for even faster
calculations.
About Us
Onyx Financial is available for either consultancy and/or software development. Please
contact
for further information.