| |
Our hedge fund product provides the backbone technology
infrastructure suitable for trading quantitative models. The user
decides which models to trade, plugs in the prime analytics for
these models and allocates capital; everything else is provided
including data connections, portfolio selections, trading decisions,
order management and reporting. The user may further develop the analytics, and procure/integrate
third-party rankings and analytics.
iBOX reduces the time to
spent on developing a new model anywhere from 50% to 80%. It
provides complete production and maintenance environment as well as
the order management system.
Quantitative Technology Infrastructure in Nutshell
A properly organized and stable quant hedge fund that outperforms
market indices, usually employs in production quite a few (minimum
3) uncorrelated models (stat-arb and fundamental). The models
undergo frequent upgrades, enhancements, and eventual replacement
adapting to ever changing market behavior. In a way, the typical
quant fund must have an assembly line for developing, deploying, and
running models.
Once the trading ideas are generated by a researcher, the actual
development cycle for stand-alone model software ranges from 6 to 12
months. Along with the innovative algorithmic jewel, there are many
more mundane and tedious tasks:
- historical and real time data acquisition
- simulation engine
- capital allocations and constraints handler
- position reconciliation and real time decision modules
- position publishing and reporting modules.
|
Based on our years of hedge fund experience, a properly architected
and designed model development framework could cut the time
necessary for introducing a new model by 80%. The overall task is
reduced to writing the algorithmic component, plugging it into the
properly configured / tuned framework, testing and running
simulations. Once deployed, the production maintenance of these
models, the risk control and capital distribution operations are
performed in a uniform and consistent manner.
Another aspect of the multi-model electronic trading environment is
the need for an integrated approach to the transaction and position
inventory across all models within the firm. While the trade
execution stations could differ reflecting specifics of various
securities and individual preferences of traders and vendors, there
are corporate level functions which require the centralized
company-wide approach at all times during the day. These include
compliance monitoring, risk control, P&L reporting, transaction
allocations and reconciliation of accounts.
Again, following our experience, a properly architected and designed
corporate order management system would be based on the integration
platform that essentially is the real time inventory caching server
equipped with the transaction messaging engine. This server would
interface with the outside brokers via its FIX protocol engine, and
internally - with all the trading stations and other necessary
applications using its messaging engine. In this manner, the hedge
fund employs best of both worlds – flexibility of competitive third
party trading stations for achieving cost-effective executions along
with the centralized corporate control for achieving right
executions and overall performance.
Both architectural components, the model development framework
(designated below as MES for Model Execution Station) and the
integration platform (designated as IS for Inventory Server)
communicate with each other, and together they constitute the
technology backbone for a successful hedge fund.
In essence, we have invented the MES approach, and by interfacing it
with IS, another generically built component, we have created a
backbone system that is unique, cost-effective and offers the
quantitative technology solution for many This backbone technology
could equally serve as the core product for either one or many hedge
funds.
As our industry experience has shown, it is generally beyond the
capabilities and resources of a single hedge fund to build such a
system completely in-house and be of commercial quality and
stability.
Model Life Cycle
This section covers all phases that any model undergoes from its
inception through the production use and the eventual replacement.
The model life cycle phases bring understanding of the requirements
for the model development tools, for the research support facilities
as well as for what it takes to run the multi-model production
environment. This understanding leads to the concept of the Model
Execution Station.
The Model life cycle consists of the following six distinct phases:
- Prototyping Primary Analytics
- Development Mode
- Research Mode
- Production Development
- Production Maintenance
- New Research
|
Quantitative Trading Infrastructure
From the top view, the Infrastructure consists of three systems:
- Input Data System - IDS
- Model Execution System (or Stations) – iBOX
- Order Management System – OMS
These systems interface via the API’s suitable for model execution
logic and performance requirements.
The Input Data
System
It delivers historical and real-time data. Some vendors establish
their databases local to the user site; some others pursue the
remote access scheme. We have introduced here the concept of the
interactive local server for selective caching. It permits using the
less expensive remote vendors yet satisfying the massive access
requirements for production needs.
The Model Execution System
Requirements to the three (3) model life cycle modes - development,
research and production – differ. Together they define functionality
of the Model Execution Station. GUI presents the advanced
functionality. An original concept of the Framework is introduced,
one that covers 80-90% of necessary code while the prime analytics
are modularized as plug-ins. This is the core point of the backbone
infrastructure; it allows dramatic reductions of the model
development time.
The different sub-systems of the iBOX provide the necessary
functionality to provide the necessary production level support for
the trading models:
The Order Management System
Hedge Fund specific inventory hierarchy and tight coupling with the
multi model input logic, make the Integration Server be a focal
point of the OMS. Such server understands the model needs, performs
FIX trading and has an open interface system for possible use of
various 3-rd party trading stations. Also, the means for trade
allocation, reconciliation and discrepancy handling must be always
considered.
The OMS market place is a very mature and competitive marketplace.
There are many to choose from. A requirement of iBOX is that the OMS
supports the ability to import and export various data set features
that iBOX in the areas of portfolio/account management and
reconciliation.
Business Model
The presented backbone infrastructure can become an business only if
the product can be leveraged between many customers. Hence, there is
a business case for a consulting service:
-
There is a core set of centrally developed and owned COMPONENTS,
and
-
In each application, there is a CUSTOMIZED SYSTEM built deploying
these components.
The top block on the Summary diagram, the core system, consists of
the following COMPONENTS:
-
MES – the Model Execution Station
-
DSS – the Data Site Server
-
IS – the Inventory Server; it is the integration platform for
storing and delivering positions and transactions. It carries and
delivers all active portfolios with their positions and transactions
updated in real-time. It also interacts with the historical
Portfolio Database (PDB) which the Inventory Server updates daily.
The latter is used for portfolio analyses and reporting.
-
AF – the Adaptor Framework that is interfaced with the IS
messaging engine. It contains the building blocks necessary for
interfacing with third party applications using in each particular
case the access API provided by the supplier of the application.
-
TC – Trade Client, a GUI application readily interfaced with the
native IS messaging system. TC provides an interface for system
testing, and in many instances it can further be used for actual
trading of the supported securities.
The CUSTOMIZED SYSTEM for a particular hedge fund consists of the
following components:
-
M 1 thru M n – a number of trading models ported to or built anew
on top of the MES framework. These models can be purchased, leased
or developed by consultants under directions of the hedge fund
researchers. Users can completely implement a model by themselves.
-
DSS – the Data Site Server, interfaced with vendor’s historical
data of choice, including price data, fundamental data, model data
(stock ranking files), corporate actions. Selections of vendor’s
data depend on the above models. Consulting services or the vendor
does interface of vendor’s data to the server.
-
IS – the Inventory Server would be installed and configured for
user’s data, accounts, supported trading practices. It is connected
to user models. Besides the installation, the particular server
configuration, its metadata definitions and interfaces to other
components are done by the consulting services as part of the custom
work.
-
Ad 1 thru Ad m – one or more adaptors implemented using the AF
tool to interface user provided trading stations and other
applications. This work would be accomplished by the consulting
services. Notice: The custom adaptors are reusable in all cases of
the same application.
-
TC – one or more Trade Clients could be installed alongside or in
place of the third party trading stations. The customer would need
no adaptors for connecting the Trade Clients.
It should be emphasized that both the IS the TC components support
not only the programmed trading from the MES based models but any
transactions for that matter entered from any interfaced trading
station (e.g. a spread sheet).
|
