The 84th Installment
The Future of Automated Trading Systems

by Hiroyuki Chishiro,
Assistant Professor, Master Program of Information Systems Architecture

The January 2016 negative interest rate policy of the Bank of Japan and the May 2016 decision of the UK to withdraw from the EU have generated a sense of financial crisis both domestically and overseas. But do you see this financial crisis as a scary thing? Or as an opportunity for investment? I am of the latter.

I develop software (OS and middleware) which targets systems with the nature of “real-time capabilities,” such as robots or automobiles, which must complete processes within a fixed amount of time. I have even had the remarkable achievement of seeing software that I researched and developed installed into robots [1]. Research will continue in the future into software for robots and automobiles, but I have also had the idea as part of my own research of applying this real-time function to automated trading systems for stocks and foreign exchange etc.

In the event of a financial crisis, manual trading produces slow decision-making, which can result in significant losses. With this in mind, while such automated trading systems are required by society, there is no de facto standard. This is why I have been researching and developing RT-Seed [2] middleware for automated trading systems. RT-Seed makes automated trading possible with real-time guarantees and improved quality of trading strategies.

I would like to introduce the “Parallel Imprecise Calculation Model” which is a feature of the RT-Seed. The Parallel Imprecise Calculation Model consists of three parts, the required part, the added part and the terminal part. The required part and the terminal part guarantee real-time performance, while the added part does not. Therefore the required part and the terminal part have higher priorities than the added part. The added part has three execution flows (completion, suspension and discarding), and each part acts independently, thus the feature of parallel processing. In the event that the added part is overloaded, by suspending its processing real-time performance of the terminal part is still guaranteed. Using the Parallel Imprecise Calculation Model, it is possible to build a flexible automated trading system for financial products.

Next, I would like to introduce an example of the Parallel Imprecise Calculation Model applied to an automated trading system. In this example, we will look at foreign exchange. The required part obtained foreign exchange data (e.g. EUR/USD) from a securities company. The added part performs technical analysis and fundamental analysis to improve the quality of the trading strategy. Technical analysis is a method of predicting future prices from past price data, and fundamental analysis is a method of predicting the appropriate price based on the soundness of the company or country. One example of technical analysis is the Bollinger Band [3]. Bollinger bands are one well-known example of technical analysis which readers with an interest in investment may have heard of. Employment is an example of fundamental analysis. The employment statistics of the United States are a particularly well-known example of employment information. The USD/JPY exchange rate tends to fluctuate rapidly around the time that US employment statistics are released. The terminal part collects the results of analysis from the added part and makes the decision as to whether to send a buy or sell order to the securities company or to take a wait-and-see approach (not trade). The required part, added part and terminal part run their processes in series, in a periodic manner. The length of the cycles can be assumed to be “seconds to minutes”.

In terms of experimenting with such an automated trading system, it was not necessary to manage my own assets. Foreign exchange can be traded for free with a demo account (an account that is traded virtually) from OANDA [4]. Free stock and economic indicators can also be obtained from Quandl [5]. RT-Seed supports the APIs provided by these companies, so it is possible for investors to experiment with their own trading strategies for free.

I have researched and developed the RT-Seed middleware in an attempt to pioneer the future of automated trading systems. It is my aim to see a day where automated trading using RT-Seed is commonplace. I hope that readers of this column have gained their own interest in automated trading systems.

[1] Takuma Shirai, Kohei Osawa, Hiroyuki Chishiro, Nobuyuki Yamasaki, and Masayuki Inaba. Design and Implementation of A High Power Robot Distributed Control System on Dependable Responsive Multithreaded Processor (D-RMTP). In Proceedings of the 4th IEEE International Conference on Cyber-Physical Systems, Networks, and Applications, pp. 19-24, October 2016.
[2] Hiroyuki Chishiro. RT-Seed: Real-Time Middleware for Semi-Fixed-Priority Scheduling. In Proceedings of the 19th IEEE International Symposium on Real-Time Computing, pp. 124-133, May 2016.
[3] John A. Bollinger, Bollinger on Bollinger Bands, 1st ed. McGraw-Hill, August 2001.
[4] OANDA．https://www.oanda.com/．
[5] Quandl．https://www.quandl.com/．