Empower Mobile Algorithmic Trading Services with Cloud Computing Junwei Ma Southwestern University of Finance and Economics Abstract: Mobile devices can offer real time connection to stock exchange market anywhere, but restricted by its limited storage and computation capability it can t do high frequent trading effectively. With the support of information technology financial market steps into millisecond age, a millisecond can differentiate between a successful trade and an unsuccessful trade so we introduce MATS (Mobile Algorithmic Trading System) to help investors optimize their investment operation against market fluctuation. Facing voluminous data flow generated by scattered exchanges and ECNs across the globe, cloud computing technology is adopted to meet demand of algorithmic trading technology. Key words: Algorithmic Trading, Mobile Finance, Cloud Computing 1. Introduction Advances in rapidly rising number of mobile devices and development of network technology, mobile devices has been changing our life habits including investment behavior. In stock market, security prices and market fluctuations may change in seconds, thus the ability to trade securities from a mobile device will enable investor keep up with any fluctuation and trade at any time to avoid risk from short term market volatility. Now you can use plenty of mobile stock trading software on Android, iphone and other mobile operation systems. These trading software can transform real time stock information and delivery trading orders to order management system of your security broker and your orders will be executed by quotation system of the stock exchange. Although mobile device offers a real time access to financial market anywhere, mobile devices restricted to hardware conditions can t match computer in processing mass data and placing orders with speed and accuracy. The small screen can t show trade information effective, touch screen is unable input trade orders quickly and precisely and the network access quality effects real time property directly. To help mobile devices users trade stocks as effective as computer platform trader, we try to incorporate algorithmic trading technology into our mobile stock trading system. Algorithmic trading is commonly defined as the use of computer algorithms to optimize trading operations. Algorithmic trading makes real time optimizing
calculation relies on mass history trading data and efficient algorithm. To facilitate technical analysis of algorithmic trading, a computing platform powerful enough to handle the mass data storage, quick retrieval and real time calculate demand is crucial. Obviously mobile devices are unequal to this task. Thus we adopt cloud computing technology to reinforce the implementation of our design, in which mobile device only send instruction and accept treated data, the data storage and computing are done by server or grid computing system. 2. Background and Related Works 2.1 Mobile Finance The significant value of mobile is arising from the convenient of mobility, Kalakota and Robinson (2001) has defined mobile commerce as e-commerce for users on the move. With adoption of information technology in finance, financial market steps in to millisecond period. Milliseconds can cost a firm an opportunity to profit from a trade (Kim 2007), thus getting market information spontaneous and real time price volatility monitoring are crucial for investors. To facility this kind of immediate trade instructions, we introduced the Mobile Algorithmic Trading System. Nearly all of the mobile finance literatures are focus on mobile bank service, but we trust with the development of e-business there will be more scholar research application in securities markets. 2.2 Algorithmic Trading Algorithmic trading (AT) is defined as the use of computer algorithms to automatically make certain trading decisions, submit orders and manage those orders after submission (Hendershott et. al.). AT has sharply increased since the implementation of decimalization in 2001. After this reform AT is accepted by both sell side and buy side to reduce transaction cost, it is thought to be responsible for about 73% of trading volume in the U.S in 2009. (Kim 2007, Hendershott 2011). Compare with traditional execution transaction orders by traders, AT has several advantages including lower commissions, reduce market impact and timing risk (Kissell et.al. 2005). If a big block order is sent to stock exchange, solely execution the trade volume would make market price move to adverse direction, thus AT splits it into several small orders and delivery them in a interval to reduce market impact intelligently Gsell (2008). 3. Design of MATS 3.1 Services Provided by MATS Our system offers three services: portfolio recommendation, securities automatic
management, algorithmic trading. Portfolio recommendation will offer advice to investors which portfolio meets their demand based on data of stock market analysis. Securities automatic management is provided by a subsystem which monitoring market fluctuation real time and make trading decision according to authorities permitted by investors. Algorithmic trading keeps all transaction orders can be implemented effectively and costless. 3.2 Architecture of MATS Mobile devices can delivery instant operation orders, but its storage ability and compute ability are far from meeting the demands of AT. Kim (2007) has cited the statistical data of Securities industry Automation Corp, message traffic of finance market is growing quickly from 56,000 messages per second in Nov. 2004 to 200,000 messages per second in Nov. 2006. The information of financial market is not only excessive, but also dispersive. The massage from one market may correlated with another, thus mobile devices are clearly unfit to give support of AT service. On the other hand, ordinary investors can t monitor stock volatility in front of a computer all day long, only mobile devices can be the medium to communicate market information and delivery orders in time any where. So we built a trading an intelligent system to connect mobile users with high frequent exchange quotation systems to facility investors optimize implementation of their trading decisions through AT technology. The MATS is represented in figure 1. User interface Mobile device: smart phone, tablet Business Layer Portfolio Recommender Subsystem Algorithmic Trading Subsystem Automatic Trading Decision Subsystem Data Layer Analyze DB Transaction DB MATS Cloud Service Stock market A Stock market B ECN A ECN B
Figure 1. Architecture of MATS Users can connect this system by mobile devices. They can offer three kinds of instructions to correspond subsystem. If it s a consultation instruction, Portfolio Recommender Subsystem will extract relevant information from two databases. Analyze database contains the recent macroeconomic data, industry analysis, and financial data of the list companies created by analyst. The consultation instrument should include investment preferences such as expected duration of this investment, risk tolerance degree, target market and so on. Then the Portfolio Recommender Subsystem will match these requirements with his trading record and analysis data, then an investment recommendation including composition of investment portfolios, holding interval, expected return and volatility are transferred to mobile device. Algorithmic Trading Subsystem plays a central role in my MATS, all trade instructions both from Automatic Trading Decision Subsystem and investors are executed by this subsystem. Since finance trade is sensitive to time delay, one more millisecond may cause an order can t be matched. The primary improvement of our trading is fast transaction direct to disadvantage of ordinary mobile stock trade transaction systems which is the time interval from spot of a trading opportunity to execution of exchange system. There are at least four network nodes of ordinary mobile transaction system: mobile devices, wireless router, order management system of broker and stock exchange system. Algorithmic trading system can connect quotation system of exchange market by DMA (direct market access) technology to access liquidity pools, this high speed access channel facilities transaction in time. Of course the Algorithmic Trading Subsystem can be used to minimize market impact and reduce transaction cost by split a block trade order into small orders then submitted them to limit book of exchange market at appropriate opportunities. All these automatic subsystems operate relies on high frequent transaction data, these data is saved in Transaction Database. The data sources are scattered globally, a list company may issue shares in different stock exchange market. For example, Bank of China has offered company s stock to public both in Shanghai Stock Exchange and Hong Kong Stock Exchange, the trade message in these two markets are closely related. Another important factor is the widely use of Electronic Communications Network(ECN is one kind of electronic communication network facilities trading outside of stock exchange where all investors post limit orders and exchange with others if their orders matched). In 2001 hundreds of ECNs were estimated to capture 37.1% of the dollar volume of trading in NASDAQ (Kim 2007). In these ECNs securities can be trade outside exchange quotation system, after American stock exchanges close, investors can transform their securities in Hong Kong or various ECNs, a global financial market forms. To meet the requirements of fast data storage, immediate message retrieval and compute under restriction of loose data source distribution and diversified mobile devices we built this system on a cloud platform.
4. Implementation 4.1 Mobile Client Side Design The mobile devices of the MATS are smart phone and tablet.basically, they can support GSM or WIFI to Access the Internet. 3G and even 4G are more welcomed, because they are more reliable and more speedy. The mobile device client application can be readily developed by JAVA ME, ADT, Object C,.NET in different mobile platforms. Client side only response for delivery of instructions and feed back operation result, the implementation of these orders are carried out in public cloud. Corresponding three subsystems there are three kinds of instructions: consultation instruction, portfolio mandate instruction and transaction instruction. The details of structure of these instructions are described in the next part. 4.2 Background Function Realization 4.2.1 Consultation Service Implementation If an investor has no idea about which securities or portfolio to selected, he can offer a recommend request to Portfolio Recommender Subsystem. This request instruction contains three factors: (i). Stock recommendation or portfolio recommendation. If an investor wants to hold a single stock, the recommendation system will selected a most qualified stock from analysis database according to other requirements. A portfolio recommendation is more complicated, investment portfolio optimizing is calculated with CAPM (capital assets pricing model) or other pricing models. (ii). Holding interval. Investor who requires liquidity may short position in next few days, so short term good trend stock is more appropriate than long term value investment portfolio. (iii). Risk bearing ability. Based on assets pricing theories, the portfolios at the portfolio efficiency frontier have different expected return and corresponding risk, a portfolio has more expected return also has larger risk so optimal portfolio is different among diverse investors. The instrument can contains more restrictions such as specified industry and specified plate, but these three factors are essential to keep the outcome accuracy. 4.2.2 Portfolio Mandate Implementation As mentioned above, Automatic Trading Decision Subsystem keeps monitoring market volatility to prevent loss caused by sudden fluctuation. Investor can set different level authority based on their personal preference. If an investor fully trusts this Automatic Trading Decision subsystem, he may give a top authority to it. Thus when a stock price drops dramatically of has significant drop trend on technology
analysis the automatic trading decision subsystem will offer corresponding transaction order to Algorithmic Trading Subsystem. The medium authority allows this subsystem trading securities after a warning message is confirmed by investor, if the user set low authority to this subsystem, it only works as an early warning system, the securities trade by investors themselves. 4.2.3 Algorithmic Trading Implementation Algorithmic Trading is the most important feature of our system, it s crucial to ensure all orders delivered to exchange effective executed. Timeliness and low transaction cost are the two main performances in the process of transaction stocks. DMA offers investors an efficient method of accessing electronic exchanges as introduced in last chapter. Algorithms typically tracks a benchmark and determine the price, quantity, opportunity and trading venues dynamically monitoring market conditions across different securities and trading venues (Hendershott 2011). These are three categories of AT strategies distinguished by their underlying benchmark. The first category of AT implemented execution strategies aims to match benchmarks generated by market itself such as volume weighted average price (VWAP). The second category of AT aims to meet benchmarks generated at the time of order submission to the algorithmic such as middle price at the moment. Third categories of AT strategies implement dynamic execution strategies as they re evaluate their strategies at each single decision time Gsell (2008). 4.3 Cloud Computing Technical Support As described above, the financial market message traffic increase rapidly and the data source is scattered all round earth, to facility rapid data storage computation and transaction, we whole system operate in cloud. The service provider develops and maintains the databases and subsystems to provide service to the mobile clients who access to the cloud. The MATS cloud services may be free or offered on a pay-per-usage model. To facilitate services we introduced above SaaS (Software as a Service) model is adopt in our cloud computing system. These cloud investors do not manage the cloud infrastructure and platform on which the application is running, this model increases elasticity of our system since developer can update algorithms real time to adapt market evolution and multi service access is available. 5. Conclusion and Further Research The current paper described a system to help investors optimize their investment behaviors against high volatility market by mobile service, algorithmic trading technology and cloud computing technology. To fit demand of mobility and easy operate, trade securities by mobile devices is a attractive choice. In our system the mobile devices connected with an intelligent supporting system, this system can offer portfolio recommendation and automatic transaction. These functions are
implemented by a modular system, investors can access to any subsystem and these subsystems can interact with each other. To cope with the challenge from rapid transaction in scattered global financial market and large amount data In the future, the primary task is realize MATS design, thus we can prove its feasibility. In the future, implementation of social network will be considered to add interaction among investors and help them to make investment decisions. Reference [1]. Kalakota R., and M. Robinson, M-Business: The Race to Mobility. McGraw-Hill, New York, 2001. [2]. Anckar B. and D Incau D. Value Creation in Mobile Commerce: Findings from a Consumer Survey. Journal of Information Technology Theory and Application. Vol. 4, Iss.1. 2002 [3]. Kim K. and Kaljuvee J. Electronic Algorithmic Trading Technology. Elsevier Inc. 2007 [4]. Hendershott T. Jones CM. and Menkveld AJ. Does Algorthmic Trading Improve Liquidity?. The Journal of Finance. Vol. 66, Iss. 1, Page 1-33. 2011 [5]. Coggins R., Lim M. and Lo K. Algorithmic Trade Execution and Market Impact, IWIF 1, Melbourne Page 518-547. 2006 [6]. Almgren R. and Lorenz J., Adaptive arrival price, Algorithmic Trading III: Precision, Control Execution. 2007 [7]. Gsell. M. Assessing the impact of algorithmic trading on markets: A simulation approach. Proceedings of the 16th European Conference on Information Systems, 2008