190 likes | 305 Views
Introduction to algo quant, an integrated trading research tool. Haksun Li haksun.li@numericalmethod.com www.numericalmethod.com. An Integrated Suite of Back Testing. Data sources Library of signals Strategy templates Sample strategies Performance measures In-sample calibration
E N D
Introduction to algo quant, an integrated trading research tool Haksun Li haksun.li@numericalmethod.com www.numericalmethod.com
An Integrated Suite of Back Testing Data sources Library of signals Strategy templates Sample strategies Performance measures In-sample calibration Out-sample back testing
An Integrated Suite Strategy Analysis Bootstrapping Customized order book Scenario analysis Auto strategy generation
Library of Components • Algo Quant is more than an application. • Algo Quant is Java library of components that you can reuse to build your own trading applications, such as: • A customized back tester • A quantitative strategy research tool • An algorithmic trading system for automatic order execution
SuanShU • Algo Quant is backed by an extensive library of numerical algorithms for building mathematical trading model. • Markov chain • Hidden Markov model • Kalman filter • Cointegration • Regression analysis
Data Sources Yahoo! Gain Capital FX rates
Data Processing • Cleaning • Extraction • Equi-time • Daily • Weekly • Filtering • Moving average
Signal Library Open-High-Low-Close (OHLC) bar Arithmetic moving average Exponential moving average RSI
Strategy Templates One of the objectives of Algo Quant is that you can prototype a quantitative trading strategy very rapidly. Reduce the time to testing out an idea. Reduce the time to production.
Message Based Ststem • Algo Quant is a message based system. • event driven • To create a strategy, you only need to handle the events that concern you. • write handlers
Signal vs. Strategy A signal takes prices (and maybe other data) to generate buy, sell signals, etc. It monitors and describes an aspect of the price process. A strategy, interacts with the market by sending orders. It determines when/what to buy and sell and how much. A strategy is a composition of signals which look at different aspects of the market.
Performance Measures P&L Max drawdown Sharpe ratio Omega Your own customized measures
Calibration • Algo Quant has a suite of optimization tools to search for optimal parameters for a strategy with respect to the (historical) data for a given objective function. • Optimizers: • mixed integer non linear programming • Objective functions: • Sharpe Ratio • Omega
Back Testing • Algo Quant is a very efficient back tester as it runs on multiple cores. • multiple set of parameters • expected P&L • variance of P&L
Customized Order Book • You can customize the way an order is handled to simulate different execution assumptions. • FIFO order book • 100% execution ratio • limit vs. market orders
composite strategy composite strategy = {simple strategies} A successful composite strategy may consist of not-so-successful strategies. A composite strategy is explainable by its constituent simple strategies. A composite strategy accounts for more market factors, hence more comprehensive.
Sample composite strategy The mean reverting strategy makes small money most of time but loses very big money on trend. The trend following strategy loses small money most of the time but makes big money on trend.
Sample composite strategy • We combine them together to form a new strategy: • run the mean reverting strategy except when there is an expected news/announcement event, e.g., NFP.
auto strategy generation a strategy a strategy a strategy a strategy a strategy a strategy a strategy a strategy a strategy a strategy a strategy a strategy search for a combination of simple strategies add the successful strategy to the pool so it becomes another simple strategy strategy verification backtester