Investment Science solutions manual 0195125177

Investment Science solutions manual 0195125177

for Investment science solutions manual contact seanjdouglas@hotmai.com

Investment Science by David Luenberger, solutions manual

Quantitative Trading System
Denis Andrey Ignatovich
denis.ignatovich@mccombs.utexa s.edu
May 5, 2006
Abstract
My interests are in the study of market microstructures, that is, how
trading takes place in the markets and how those markets are organized.
Models are designed to describe aspects of these organizations and one
needs flexible toolsets for model description and performance analysis.
The current step in my research is an implementation of a quantitative
trading system. Not only is this a challenging systems engineering project,
but also a powerful mechanism for data analysis and trade algorithm
description.
Trading Model is defined as an investment tool comprised of buy and
sell recommendations. These recommendations are much more complex
than simple price change forecasts. They must be taylored to the investor
in terms of his/her risk exposure, past trading history, and the market
microstructure with its own constraints on the trade execution. A trading
model has three main components:
• Generation of the trading recommendations.
• Accounting of the simulated transactions and their price impacts.
• Generation of the model statistics by the performance calculator.
A Trading System is, in turn, an environment where users define and,
through execution feedback, adjust their trading models. An essential
part of the system is the user-interface: it must be eloquent enough to
allow an ease-of-use and, at the same time, powerful to describe most
sophisticated trading algorithms. As in forecasting or other applications,
trading models rely heavily on the quality of financial data. This constraint
on the Trading System, a supply of the tick-by-tick data, is just an
example of the multitude of requirements of a functional real-time trading
environment.
Thie following paper describes an implementation of a quantiative
trading system designed to incorporate features representative of a commercial
grade trading environment. The system that I propose includes
programmatic access to the underlying execution framework, powerful
Python-based algorithm description environment, real-time data support,
and mathetmatical interfaces, including support for ARCH-based equity
volatility models.
This paper, in part, fulfills the degree requirement for Bachelor of
Science in Computer Sciences (Turing Scholars Option).
1

Quant is an algorithmic trading system. Using Python, users define data
analysis and trading algorithms. The destinction is that former operate on inhomogeneous
time series (tick data) and publish results to trading algorithms,
that make trade decisions for portfolios they are responsible for. Users create
portfolios by specifying their descriptions and initial capital. A portfolio may be
managed by the PortfolioManager (part of TradingModel, please see below) that
simply updates portfolio positions with the user manually modifying their composition
weights. Alternatively, the user will assign a custom trading algorithm
that would make the decisions automatically upon receiving signals of new data
arrival. R-statistical environment provides the many necessary computational
tools for making informed decisions.
2
2 System Model
QuantWorld’s synthetic trades Designing solution to an engineering problem
must start with analyzing the requirements in terms of input data, precision/
algorithm running complexity, and results’ presentation. In the case of
Quant, there is an important issue of quality data: real-world trading systems
rely on fast and precise trade information (and recently with the introduction
of NYSE’s OpenBook, even the specialists’ books) for decision making. Lack of
access to this level of data forces numerous assumptions on the part of the trader
designing the algorithms. Yahoo! Finance provides 20-minute delayed price information
distributed once per minute. It is the most commonly used source of
financial data outside of the commercial realm, and due to the economic constraints
of this project, it will service us with the foundation of our trade data.
The necessary environment for the trade algorithm execution must provide the
most up-to-date price information. The transition from Yahoo! Finance data
to our algorithm environment (TradingModel ) is supplied by QuantWorld, a
component outside the trading system that simulates high-frequency data generation.
The simulation will result in synthetic ticks containing bid/ask, price,
and volume information distributed according to the parameters set by the user.
The necessity of this price generation is specific to the author’s use of the system,
and therefore QuantWorld’s residence outside of the system allows for an
substitution of a pricing information feed accomodating purposes of another
user.
Data vs. Trading Algorithms Many operations performed by the users’
algorithms would be quiet common data handling tasks. Thus, the decision to
provide two algorithmic interfaces was raised by the need to factor out redundant
tasks. The Matrix runs data-analysis algorithms and publishes the results for
the trading algorithms.
UserInterface The UserInterface handles the task of informing