Design And Implementation Of Programmed Trading Strategy Backtesting System

With the continuous development of global economy and the advancement of internationalization of the RMB,financial industry has put forward higher requirements for transaction costs and transaction efficiency,and electronic trading platforms relying on modern computer technology have emerged as the times require.However,the main functions of current mainstream electronic trading platforms are still order executing and transaction filing,and higher-level decision-making behaviors are still done manually by human traders.With most of the participants in the market completing the electronic trading platform construction,traders urgently need a programmatic trading strategy development platform based on the decision-making level.In view of this,the paper designs a programmatic trading strategy backtesting system based on an event-driven model,aiming to save labor costs and to improve the efficiency of trading strategy research and development.Compared with traditional backtesting system,the backtesting engine implemented in the system is based on an event-driven model: callback functions are designed according to the upstream and downstream interfaces and order status in electronic trading provided by the China Foreign Exchange Trade System(CFETS),simulating real trading process of the market and improving reliability of trading strategies backtesting results.The backtesting engine also simulates the decision making process of the exchange by providing two matching methods: bar data matching and market data matching,using real historical data to simulate trading results to the strategy logic,therefore improve the precision of backtesting process for the trading strategy.The backtesting service is provided by the Hadoop cluster.It has the ability to use historical data tick by tick on strategy backtesting for a large period of time,which saves local computing and storage resources and facilitates the upgrade and expansion of subsequent functions of the system.The backtesting system adopts B/S architecture,which can be accessed by users through web pages and not need to download client.The front-end part adopts the React.js framework popular in the industry,and by component designing reduces the complexity of functional coupling and ensures the stability of the system.The back-end adopts the Django framework in which APIs are designed based on RESTful to standardize data transmission process.Distributed storage of historical data and system files are provided by HDFS,and strategy running process is implemented by Spark,which retains advantages of distributed computing and avoids high load of managing nodes.Finally,the system realizes user management module,strategy management module,data management module,process management module,backtesting result module and system management module,and passes functional and non-functional tests.