The Penn-Lehman Automated Trading Project

IFIP International Federation for Information Processing, 2006

Three core technologies are needed for automated trading: data mining, intelligent trading agents and virtual institutions in which informed trading agents can trade securely both with each other and with human agents in a natural way. This paper describes a demonstrable prototype that integrates these three technologies and is available on the World Wide Web. This is part of a larger project that aims to make informed automated trading a reality.

2015

2019

The goal of this Interactive Qualifying Project (IQP) is to test whether a diverse system of automatic trading strategies can be profitable, ie., have a higher return than market indexes after costs. The second objective is to understand how a system of systems can improve return rates compared to a single system. In order to do this each member of the team has developed their own trading strategy for a specific market (Stocks and Forex). Each set of strategies will have rules for entry, exit and position sizing according to its unique trading patterns.

2012

ajor advancements in artificial intelligence have been stimulated by well-designed competitions that tackle intriguing and often very complex problems. Chess, poker, stock trading, real-time strategy games, robot soccer, robot rescue or planning, and autonomous vehicles are among the most well known. Adaptability, proactiveness, and interoperability are essential characteristics of these games. Many of these competitions focus on development of rational behavior that would allow dominance over competitors. Apart from game play, though, one of the main reasons for these competitions is to bridge the gap between AI and real-life domains. In many real-life domains, such as trading environments, selfinterested entities need to operate subject to limited time and information. Additionally, the web has mediated an ever broader range of transactions, urging participants to concurrently trade across multiple markets. All these have generated the need for technologies that empower prompt investigation of large volumes of data and rapid evaluation of numerous alternative strategies in the face of constantly changing market conditions (Bichler, Gupta, and Ketter 2010). AI and machine-learning techniques, including neural networks and genetic algorithms, are continuously gaining ground in the support of such trading scenarios. User modeling, price forecasting, market equilibrium prediction, and strategy optimization are typical cases where AI typically provides reliable solutions. Yet, the adoption and deployment of AI practices in real trading environments remains limited, since the proprietary nature of markets precludes open benchmarking, which is critical for further scientific progress. To Competition Reports

Market making (MM) strategies have played an important role in the electronic stock market. However, the MM strategies without any forecasting power are not safe while trading. In this paper, we design and implement a twotier framework, which includes a trading signal generator based on a supervised learning approach and an event-driven MM strategy. The proposed generator incorporates the information within order book microstructure and market news to provide directional predictions. The MM strategy in the second tier trades on the signals and prevents itself from profit loss led by market trending. Using half a year price tick data from Tokyo Stock Exchange (TSE) and Shanghai Stock Exchange (SSE), and corresponding Thomson Reuters news of the same time period, we conduct the back-testing and simulation on an industrial near-to-reality simulator. From the empirical results, we find that 1) strategies with signals perform better than strategies without any signal in terms of average daily profit and loss (PnL) and sharpe ratio (SR), and 2) correct predictions do help MM strategies readjust their quoting along with market trending, which avoids the strategies triggering stop loss procedure that further realizes the paper loss.

2009

After exchanges and alternative trading venues have introduced electronic execution mechanisms worldwide, the focus of the securities trading industry shifted to the use of fully electronic trading engines by banks, brokers and their institutional customers. These Algorithmic Trading engines enable order submissions without human intervention based on quantitative models applying historical and real-time market data. Although there is a widespread discussion on the pros and cons of Algorithmic Trading and on its impact on market volatility and market quality, little is known on how algorithms actually place their orders in the market and whether and in which respect this differs form other order submissions. Based on a dataset that-for the first time-includes a specific flag to enable the identification of orders submitted by Algorithmic Trading engines, the paper investigates the extent of Algorithmic Trading activity and specifically their order placement strategies in comparison to human traders in the Xetra trading system. It is shown that Algorithmic Trading has become a relevant part of overall market activity and that Algorithmic Trading engines fundamentally differ from human traders in their order submission, modification and deletion behavior as they exploit real-time market data and latest market movements.

Operations Research Proceedings 2008, 2009

Investors which trade in financial markets are interested in buying at low and selling at high prices. We suggest to solve this type of problem with an online algorithm. This active trading algorithm is based on reservation prices. The effectiveness of the algorithm is analyzed from a worst case and an average case point of view. We also compare the average case and the worst case bounds using simulation on historical data. Moreover, we want to give an answer to the question if the suggested active online trading algorithm shows a superior behaviour to buy-and-hold policies.

PACIS 2005 …, 2005

A global trend towards automated trading systems encourages the appearances of different kinds of decision support systems (DSS) for securities trading. However, the efficiency and effectiveness of the trading decisions and investment strategies suggested by a computer based system are always being oppugned by researchers for abnegating the specialists' professional intermediary roles in traditional floor trading. Without proper validations, evaluations and comparisons, those valuable decision support systems can hardly be adopted trustily. This paper presents a Web-service-agents-based Securities Trading Simulation System (STSS) as a platform to validate, evaluate and compare the existing securities trading decision support systems by using the real world securities trading data from Chicago Stock Exchange. Agent technology is applied to deal with the complex, dynamic, and distributed securities trading processes. Web-services techniques are proposed for more interoperability and scalability in network-based business environment. By integrating agent technology with Web-services to make use of the advantages of both, this approach provides a more intelligent, flexible, and comprehensive platform to simulate the securities trading process. Performance of different decision support systems for securities trading can be validated, evaluated and compared either separately or simultaneously on a real time basis.

Some major financial markets are currently reporting that 50% or more of all transactions are now executed by automated trading systems (ATS). To understand the impact of ATS proliferation on the global financial markets, academic studies often use standard reference strategies, such as "AA" and "ZIP", to model the behaviour of real trading systems. Disturbingly, we show that the reference algorithms presented in the literature are ambiguous, thus reducing the validity of strict comparative studies. As a remedy, we suggest disambiguated standard implementations of AA and ZIP. Using Exchange Portal (ExPo), an open-source financial exchange simulation platform designed for realtime behavioural economic experiments involving human traders and/or trader-agents, we study the e↵ects of disambiguating AA and ZIP, before introducing a novel method of assignment-adaptation (ASAD). Experiments show that introducing ASAD agents into a market with shocks can produce counter-intuitive market dynamics.

2013

This book serves two purposes. First, it teaches the importance of using sophisticated yet accessible statistical methods to evaluate a trading system before it is put to real-world use. In order to accommodate readers having limited mathematical background, these techniques are illustrated with step-by-step examples using actual market data, and all examples are explained in plain language. Second, this book shows how the free program TSSB (Trading System Synthesis & Boosting) can be used to develop and test trading systems. The machine learning and statistical algorithms available in TSSB go far beyond those available in other off-the-shelf development software. Intelligent use of these state-of-the-art techniques greatly improves the likelihood of obtaining a trading system whose impressive backtest results continue when the system is put to use in a trading account. Among other things, this book will teach the reader how to: Estimate future performance with rigorous algorithms E...