M.Sc. / Advancements in technology used in financial markets have led to substantial automation of tasks within the financial industry. Data analysis, trade execution and trade processing have been automated, reducing costs and increasing productivity. Algorithmic trading is the automated execution of trades on an electronic trading platform; it has been used to gain competitive advantage in financial markets since the early 1990s. Algorithmic trading applications, which must analyse information and determine whether to buy or sell, are well suited to the use of autonomous software agents. Multi-agent systems are better suited to the increasing complexity of algorithmic trading systems and the flexibility required by rapidly changing markets than single-agent systems. The granularity of components (agents) in multi-agent systems also promotes reuse and simplifies individual agent design. Algorithmic trading is, however, subject to challenges specifically in terms of data volume, speed of access and speed of processing. In order to utilise a multi-agent system solution the interactions between agents which allow distributed problem solving must be as efficient as possible. This dissertation investigates the use of indirect coordination to improve the efficiency of interactions between agents in multi-agent systems and to simplify agent design. Indirect coordination utilises environment abstractions known as artefacts to facilitate interaction between agents; such interaction can be simple data transfer or requests, complex coordination protocols as well as negotiation protocols. The investigation resulted in a framework that allows agents to transition between direct and indirect interaction techniques based on the specific interaction task at hand. The framework is built on two existing platforms, ii Java Agent DEvelopment Framework (JADE) and Common ARTifact Infrastructure for AGents Open environments (CARTAGO). These platforms are combined into the JADE-CARTAGO Algorithmic Trading (JCAT) framework that provides the infrastructure needed for both direct and indirect interactions. Investigations into the performance of the JCAT framework have shown that artefacts improve interaction efficiency by reducing data loss in tasks such as information publishing, and perform as well as direct communication within certain constraints for other tasks. When limiting the number of agents in an interaction to 50 agents, artefacts perform at least as well as direct communication using agent communication language messages.

The research at hand aims to define effectiveness of algorithmic trading, comparing with different benchmarks represented by several types of indexes. How big returns can be gotten by algorithmic trading, taking into account the costs of informational and trading infrastructure needed for robot trading implementation? To get the result, it’s necessary to compare two opposite trading strategies: 1) Algorithmic trading (implemented by high-frequency trading robot (based on statistic arbitrage strategy) and trend-following trading robot (based on the indicator Exponential Moving Average with the Variable Factor of Smoothing)) 2) Index investing strategy (classical index strategies “buy and hold”, implemented by four different types of indexes: Capitalization weight index, Fundamental indexing, Equal-weighted indexing, Risk-based indexation/minimal variance). According to the results, it was found that at the current phase of markets’ development, it is theoretically possible for algorithmic trading (and especially high-frequency strategies) to exceed the returns of index strategy, but we should note two important factors: 1) Taking into account all of the costs of organization of high-frequency trading (brokerage and stock exchanges commissions, trade-related infrastructure maintenance, etc.), the difference in returns (with superiority of high-frequency strategy) will be much less . 2) Given the fact that “markets’ efficiency” is growing every year (see more about it further in thesis), and the returns of high-frequency strategies tends to decrease with time (see more about it further in thesis), it is quite logical to assume that it will be necessary to invest more and more in trading infrastructure to “fix” the returns of high-frequency trading strategies on a higher level, than the results of index investing strategies.

No description available.

Building automated trading systems has long been one of the most cutting-edge and exciting fields in the financial industry. In this research project, we built a trading system based on machine learning methods. We used the Recurrent Reinforcement Learning (RRL) algorithm as our fundamental algorithm, and by introducing Genetic Algorithms (GA) in the optimization procedure, we tackled the problems of picking good initial values of parameters and dynamically updating the learning speed in the original RRL algorithm. We call this optimization algorithm the Evolutionary Recurrent Reinforcement Learning algorithm (ERRL), or the GA-RRL algorithm. ERRL allows us to find many local optimal solutions easier and faster than the original RRL algorithm. Finally, we implemented the GA-RRL system on EUR/USD at a 5-minute level, and the backtest performance showed that our GA-RRL system has potentially promising profitability. In future research we plan to introduce some risk control mechanism, implement the system on different markets and assets, and perform backtest at higher frequency level.

Thesis (M.M. (Finance & Investment))--University of the Witwatersrand, Faculty of Commerce, Law and Management, Graduate School of Business Administration, 2013. / The evidence put forward by Zhang (2010) indicates that algorithmic trading can potentially generate the momentum effect evident in empirical market research. In addition, upon analysis of the literature, it is apparent that algorithmic traders possess a comparative informational advantage relative to regular traders. Finally, the theoretical model proposed by Wang (1993), indicates that the informational differences between traders fundamentally influences the nature of asset prices, even generating serial return correlations. Thus, applied to the study, the theory holds that algorithmic trading would have a significant effect on security return dynamics, possibly even engendering the momentum effect. This paper tests such implications by proposing a theory to explain the momentum effect based on the hypothesis that algorithmic traders possess Innovative Information about a firm’s future performance. From this perspective, Innovative Information can be defined as the information derived from the ability to accumulate, differentiate, estimate, analyze and utilize colossal quantities of data by means of adept techniques, sophisticated platforms, capabilities and processing power. Accordingly, an algorithmic trader’s access to various complex computational techniques, infrastructure and processing power, together with the constraints to human information processing, allow them to make judgments that are superior to the judgments of other traders. This particular aspect of algorithmic trading remains, to the best of my knowledge, unexplored as an avenue or mechanism, through which algorithmic trading could possibly affect the momentum effect and thus market efficiency. Interestingly, by incorporating this information variable into a simplified representative agent model, we are able to produce return patterns consistent with the momentum effect in its entirety. The general thrust of our results, therefore, is that algorithmic trading can hypothetically generate the return anomaly known as the momentum effect. Our results give credence to the assumption that algorithmic trading is having a detrimental effect on stock market efficiency.

<p>This thesis looks into how one could use Hierarchal Temporal Memory (HTM) networks to generate models that could be used as trading algorithms. The thesis begins with a brief introduction to algorithmic trading and commonly used concepts when developing trading algorithms. The thesis then proceeds to explain what an HTM is and how it works. To explore whether an HTM could be used to generate models that could be used as trading algorithms, the thesis conducts a series of experiments. The goal of the experiments is to iteratively optimize the settings for an HTM and try to generate a model that when used as a trading algorithm would have more profitable trades than losing trades. The setup of the experiments is to train an HTM to predict if it is a good time to buy some shares in a security and hold them for a fixed time before selling them again. A fair amount of the models generated during the experiments was profitable on data the model have never seen before, therefore the author concludes that it is possible to train an HTM so it can be used as a profitable trading algorithm.</p>

This thesis looks into how one could use Hierarchal Temporal Memory (HTM) networks to generate models that could be used as trading algorithms. The thesis begins with a brief introduction to algorithmic trading and commonly used concepts when developing trading algorithms. The thesis then proceeds to explain what an HTM is and how it works. To explore whether an HTM could be used to generate models that could be used as trading algorithms, the thesis conducts a series of experiments. The goal of the experiments is to iteratively optimize the settings for an HTM and try to generate a model that when used as a trading algorithm would have more profitable trades than losing trades. The setup of the experiments is to train an HTM to predict if it is a good time to buy some shares in a security and hold them for a fixed time before selling them again. A fair amount of the models generated during the experiments was profitable on data the model have never seen before, therefore the author concludes that it is possible to train an HTM so it can be used as a profitable trading algorithm.

This thesis introduces a selection of models for optimal execution of financial assets at the tactical level. As opposed to optimal scheduling, which defines a trading schedule for the trader, this thesis investigates how the trader should interact with the order book. If a trader is aggressive he will execute his order using market orders, which will negatively feedback on his execution price through market impact. Alternatively, the models we focus on consider a passive trader who places limit orders into the limit-order book and waits for these orders to be filled by market orders from other traders. We assume these models do not exhibit market impact. However, given we await market orders from other participants to fill our limit orders a new risk is borne: execution risk. We begin with an extension of Guéant et al. (2012b) who through the use of an exponential utility, standard Brownian motion, and an absolute decay parameter were able to cleverly build symmetry into their model which significantly reduced the complexity. Our model consists of geometric Brownian motion (and mean-reverting processes) for the asset price, a proportional control parameter (the additional amount we ask for the asset), and a proportional decay parameter, implying that the symmetry found in Guéant et al. (2012b) no longer exists. This novel combination results in asset-dependent trading strategies, which to our knowledge is a unique concept in this framework of literature. Detailed asymptotic analyses, coupled with advanced numerical techniques (informing the asymptotics) are exploited to extract the relevant dynamics, before looking at further extensions using similar methods. We examine our above mentioned framework, as well as that of Guéant et al. (2012), for a trader who has a basket of correlated assets to liquidate. This leads to a higher-dimensional model which increases the complexity of both numerically solving the problem and asymptotically examining it. The solutions we present are of interest, and comparable with Markowitz portfolio theory. We return to our framework of a single underlying and consider four extensions: a stochastic volatility model which results in an added dimension to the problem, a constrained optimisation problem in which the control has an explicit lower bound, changing the exponential intensity to a power intensity which results in a reformulation as a singular stochastic control problem, and allowing the trader to trade using both market orders and limit orders resulting in a free-boundary problem. We complete the study with an empirical analysis using limit-order book data which contains multiple levels of the book. This involves a novel calibration of the intensity functions which represent the limit-order book, before backtesting and analysing the performance of the strategies.

No / The article offers information on expansion of Artificial Intelligence (AI) in the financial services industry. Topics include Financial institutions see in it more opportunities for efficiency generation, improved profitability, and opportunities for differentiation for the building of competitive advantages; and develop, to improve reporting, and compliance processes.

Die Elektronisierung der Finanzmärkte ist in den letzten Jahren weit vorangeschritten. Praktisch jede Börse verfügt über ein elektronisches Handelssystem. In diesem Kontext beschreibt der Begriff Algorithmic Trading ein Phänomen, bei dem Computerprogramme den Menschen im Wertpapierhandel ersetzen. Sie helfen dabei Investmententscheidungen zu treffen oder Transaktionen durchzuführen. Algorithmic Trading selbst ist dabei nur eine unter vielen Innovationen, welche die Entwicklung des Börsenhandels geprägt haben. Hier sind z.B. die Erfindung der Telegraphie, des Telefons, des FAX oder der elektronische Wertpapierabwicklung zu nennen. Die Frage ist heute nicht mehr, ob Computerprogramme im Börsenhandel eingesetzt werden. Sondern die Frage ist, wo die Grenze zwischen vollautomatischem Börsenhandel (durch Computer) und manuellem Börsenhandel (von Menschen) verläuft. Bei der Erforschung von Algorithmic Trading wird die Wissenschaft mit dem Problem konfrontiert, dass keinerlei Informationen über diese Computerprogramme zugänglich sind. Die Idee dieser Dissertation bestand darin, dieses Problem zu umgehen und Informationen über Algorithmic Trading indirekt aus der Analyse von (Fonds-)Renditen zu extrahieren. Johannes Gomolka untersucht daher die Forschungsfrage, ob sich Aussagen über computergesteuerten Wertpapierhandel (kurz: Algorithmic Trading) aus der Analyse von (Fonds-)Renditen ziehen lassen. Zur Beantwortung dieser Forschungsfrage formuliert der Autor eine neue Definition von Algorithmic Trading und unterscheidet mit Buy-Side und Sell-Side Algorithmic Trading zwei grundlegende Funktionen der Computerprogramme (die Entscheidungs- und die Transaktionsunterstützung). Für seine empirische Untersuchung greift Gomolka auf das Multifaktorenmodell zur Style-Analyse von Fung und Hsieh (1997) zurück. Mit Hilfe dieses Modells ist es möglich, die Zeitreihen von Fondsrenditen in interpretierbare Grundbestandteile zu zerlegen und den einzelnen Regressionsfaktoren eine inhaltliche Bedeutung zuzuordnen. Die Ergebnisse dieser Dissertation zeigen, dass man mit Hilfe der Style-Analyse Aussagen über Algorithmic Trading aus der Analyse von (Fonds-)Renditen machen kann. Die Aussagen sind jedoch keiner technischen Natur, sondern auf die Analyse von Handelsstrategien (Investment-Styles) begrenzt. / During the last decade the electronic trading on the stock exchanges advanced rapidly. Today almost every exchange is running an electronic trading system. In this context the term algorithmic trading describes a phenomenon, where computer programs are replacing the human trader, when making investment decisions or facilitating transactions. Algorithmic trading itself stands in a row of many other innovations that helped to develop the financial markets technologically (see for example telegraphy, the telephone, FAX or electronic settlement). Today the question is not, whether computer programs are used or not. The question arising is rather, where the border between automatic, computer driven and human trading can be drawn. Conducting research on algorithmic trading confronts scientists always with the problem of limited availability of information. The idea of this dissertation is to circumnavigate this problem and to extract information indirectly from an analysis of a time series of (fund)-returns data. The research question here is: Is it possible to draw conclusions about algorithmic trading from an analysis of (funds-)return data? To answer this question, the author develops a complete definition of algorithmic trading. He differentiates between Buy-Side and Sell-Side algorithmic trading, depending on the functions of the computer programs (supporting investment-decisions or transaction management). Further, the author applies the multifactor model of the style analysis, formely introduced by Fung and Hsieh (1997). The multifactor model allows to separate fund returns into regression factors that can be attributed to different reasons. The results of this dissertation do show that it is possible to draw conclusions about algorithmic trading out of the analysis of funds returns. Yet these conclusions cannot be of technical nature. They rather have to be attributed to investment strategies (investment styles).